File size: 25,003 Bytes
28958dc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
/******************************************************************************
 * Copyright (c) 2011, Duane Merrill.  All rights reserved.
 * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
 * 
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:
 *     * Redistributions of source code must retain the above copyright
 *       notice, this list of conditions and the following disclaimer.
 *     * Redistributions in binary form must reproduce the above copyright
 *       notice, this list of conditions and the following disclaimer in the
 *       documentation and/or other materials provided with the distribution.
 *     * Neither the name of the NVIDIA CORPORATION nor the
 *       names of its contributors may be used to endorse or promote products
 *       derived from this software without specific prior written permission.
 * 
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
 * DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
 * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 ******************************************************************************/

/**
 * \file
 * The cub::BlockDiscontinuity class provides [<em>collective</em>](index.html#sec0) methods for flagging discontinuities within an ordered set of items partitioned across a CUDA thread block.
 */

#pragma once

#include "../config.cuh"
#include "../util_type.cuh"
#include "../util_ptx.cuh"

/// Optional outer namespace(s)
CUB_NS_PREFIX

/// CUB namespace
namespace cub {

template <
    typename    T,
    int         BLOCK_DIM_X,
    int         BLOCK_DIM_Y     = 1,
    int         BLOCK_DIM_Z     = 1,
    int         PTX_ARCH        = CUB_PTX_ARCH>
class BlockAdjacentDifference
{
private:

    /******************************************************************************
     * Constants and type definitions
     ******************************************************************************/

    /// Constants
    enum
    {
        /// The thread block size in threads
        BLOCK_THREADS = BLOCK_DIM_X * BLOCK_DIM_Y * BLOCK_DIM_Z,
    };


    /// Shared memory storage layout type (last element from each thread's input)
    struct _TempStorage
    {
        T first_items[BLOCK_THREADS];
        T last_items[BLOCK_THREADS];
    };


    /******************************************************************************
     * Utility methods
     ******************************************************************************/

    /// Internal storage allocator
    __device__ __forceinline__ _TempStorage& PrivateStorage()
    {
        __shared__ _TempStorage private_storage;
        return private_storage;
    }


    /// Specialization for when FlagOp has third index param
    template <typename FlagOp, bool HAS_PARAM = BinaryOpHasIdxParam<T, FlagOp>::HAS_PARAM>
    struct ApplyOp
    {
        // Apply flag operator
        static __device__ __forceinline__ T FlagT(FlagOp flag_op, const T &a, const T &b, int idx)
        {
            return flag_op(b, a, idx);
        }
    };

    /// Specialization for when FlagOp does not have a third index param
    template <typename FlagOp>
    struct ApplyOp<FlagOp, false>
    {
        // Apply flag operator
        static __device__ __forceinline__ T FlagT(FlagOp flag_op, const T &a, const T &b, int /*idx*/)
        {
            return flag_op(b, a);
        }
    };

    /// Templated unrolling of item comparison (inductive case)
    template <int ITERATION, int MAX_ITERATIONS>
    struct Iterate
    {
        // Head flags
        template <
            int             ITEMS_PER_THREAD,
            typename        FlagT,
            typename        FlagOp>
        static __device__ __forceinline__ void FlagHeads(
            int                     linear_tid,
            FlagT                   (&flags)[ITEMS_PER_THREAD],         ///< [out] Calling thread's discontinuity head_flags
            T                       (&input)[ITEMS_PER_THREAD],         ///< [in] Calling thread's input items
            T                       (&preds)[ITEMS_PER_THREAD],         ///< [out] Calling thread's predecessor items
            FlagOp                  flag_op)                            ///< [in] Binary boolean flag predicate
        {
            preds[ITERATION] = input[ITERATION - 1];

            flags[ITERATION] = ApplyOp<FlagOp>::FlagT(
                flag_op,
                preds[ITERATION],
                input[ITERATION],
                (linear_tid * ITEMS_PER_THREAD) + ITERATION);

            Iterate<ITERATION + 1, MAX_ITERATIONS>::FlagHeads(linear_tid, flags, input, preds, flag_op);
        }

        // Tail flags
        template <
            int             ITEMS_PER_THREAD,
            typename        FlagT,
            typename        FlagOp>
        static __device__ __forceinline__ void FlagTails(
            int                     linear_tid,
            FlagT                   (&flags)[ITEMS_PER_THREAD],         ///< [out] Calling thread's discontinuity head_flags
            T                       (&input)[ITEMS_PER_THREAD],         ///< [in] Calling thread's input items
            FlagOp                  flag_op)                            ///< [in] Binary boolean flag predicate
        {
            flags[ITERATION] = ApplyOp<FlagOp>::FlagT(
                flag_op,
                input[ITERATION],
                input[ITERATION + 1],
                (linear_tid * ITEMS_PER_THREAD) + ITERATION + 1);

            Iterate<ITERATION + 1, MAX_ITERATIONS>::FlagTails(linear_tid, flags, input, flag_op);
        }

    };

    /// Templated unrolling of item comparison (termination case)
    template <int MAX_ITERATIONS>
    struct Iterate<MAX_ITERATIONS, MAX_ITERATIONS>
    {
        // Head flags
        template <
            int             ITEMS_PER_THREAD,
            typename        FlagT,
            typename        FlagOp>
        static __device__ __forceinline__ void FlagHeads(
            int                     /*linear_tid*/,
            FlagT                   (&/*flags*/)[ITEMS_PER_THREAD],         ///< [out] Calling thread's discontinuity head_flags
            T                       (&/*input*/)[ITEMS_PER_THREAD],         ///< [in] Calling thread's input items
            T                       (&/*preds*/)[ITEMS_PER_THREAD],         ///< [out] Calling thread's predecessor items
            FlagOp                  /*flag_op*/)                            ///< [in] Binary boolean flag predicate
        {}

        // Tail flags
        template <
            int             ITEMS_PER_THREAD,
            typename        FlagT,
            typename        FlagOp>
        static __device__ __forceinline__ void FlagTails(
            int                     /*linear_tid*/,
            FlagT                   (&/*flags*/)[ITEMS_PER_THREAD],         ///< [out] Calling thread's discontinuity head_flags
            T                       (&/*input*/)[ITEMS_PER_THREAD],         ///< [in] Calling thread's input items
            FlagOp                  /*flag_op*/)                            ///< [in] Binary boolean flag predicate
        {}
    };


    /******************************************************************************
     * Thread fields
     ******************************************************************************/

    /// Shared storage reference
    _TempStorage &temp_storage;

    /// Linear thread-id
    unsigned int linear_tid;


public:

    /// \smemstorage{BlockDiscontinuity}
    struct TempStorage : Uninitialized<_TempStorage> {};


    /******************************************************************//**
     * \name Collective constructors
     *********************************************************************/
    //@{

    /**
     * \brief Collective constructor using a private static allocation of shared memory as temporary storage.
     */
    __device__ __forceinline__ BlockAdjacentDifference()
    :
        temp_storage(PrivateStorage()),
        linear_tid(RowMajorTid(BLOCK_DIM_X, BLOCK_DIM_Y, BLOCK_DIM_Z))
    {}


    /**
     * \brief Collective constructor using the specified memory allocation as temporary storage.
     */
    __device__ __forceinline__ BlockAdjacentDifference(
        TempStorage &temp_storage)  ///< [in] Reference to memory allocation having layout type TempStorage
    :
        temp_storage(temp_storage.Alias()),
        linear_tid(RowMajorTid(BLOCK_DIM_X, BLOCK_DIM_Y, BLOCK_DIM_Z))
    {}


    //@}  end member group
    /******************************************************************//**
     * \name Head flag operations
     *********************************************************************/
    //@{


#ifndef DOXYGEN_SHOULD_SKIP_THIS    // Do not document

    template <
        int             ITEMS_PER_THREAD,
        typename        FlagT,
        typename        FlagOp>
    __device__ __forceinline__ void FlagHeads(
        FlagT           (&head_flags)[ITEMS_PER_THREAD],    ///< [out] Calling thread's discontinuity head_flags
        T               (&input)[ITEMS_PER_THREAD],         ///< [in] Calling thread's input items
        T               (&preds)[ITEMS_PER_THREAD],         ///< [out] Calling thread's predecessor items
        FlagOp          flag_op)                            ///< [in] Binary boolean flag predicate
    {
        // Share last item
        temp_storage.last_items[linear_tid] = input[ITEMS_PER_THREAD - 1];

        CTA_SYNC();

        if (linear_tid == 0)
        {
            // Set flag for first thread-item (preds[0] is undefined)
            head_flags[0] = 1;
        }
        else
        {
            preds[0] = temp_storage.last_items[linear_tid - 1];
            head_flags[0] = ApplyOp<FlagOp>::FlagT(flag_op, preds[0], input[0], linear_tid * ITEMS_PER_THREAD);
        }

        // Set head_flags for remaining items
        Iterate<1, ITEMS_PER_THREAD>::FlagHeads(linear_tid, head_flags, input, preds, flag_op);
    }

    template <
        int             ITEMS_PER_THREAD,
        typename        FlagT,
        typename        FlagOp>
    __device__ __forceinline__ void FlagHeads(
        FlagT           (&head_flags)[ITEMS_PER_THREAD],    ///< [out] Calling thread's discontinuity head_flags
        T               (&input)[ITEMS_PER_THREAD],         ///< [in] Calling thread's input items
        T               (&preds)[ITEMS_PER_THREAD],         ///< [out] Calling thread's predecessor items
        FlagOp          flag_op,                            ///< [in] Binary boolean flag predicate
        T               tile_predecessor_item)              ///< [in] <b>[<em>thread</em><sub>0</sub> only]</b> Item with which to compare the first tile item (<tt>input<sub>0</sub></tt> from <em>thread</em><sub>0</sub>).
    {
        // Share last item
        temp_storage.last_items[linear_tid] = input[ITEMS_PER_THREAD - 1];

        CTA_SYNC();

        // Set flag for first thread-item
        preds[0] = (linear_tid == 0) ?
            tile_predecessor_item :              // First thread
            temp_storage.last_items[linear_tid - 1];

        head_flags[0] = ApplyOp<FlagOp>::FlagT(flag_op, preds[0], input[0], linear_tid * ITEMS_PER_THREAD);

        // Set head_flags for remaining items
        Iterate<1, ITEMS_PER_THREAD>::FlagHeads(linear_tid, head_flags, input, preds, flag_op);
    }

#endif // DOXYGEN_SHOULD_SKIP_THIS


    template <
        int             ITEMS_PER_THREAD,
        typename        FlagT,
        typename        FlagOp>
    __device__ __forceinline__ void FlagHeads(
        FlagT           (&head_flags)[ITEMS_PER_THREAD],    ///< [out] Calling thread's discontinuity head_flags
        T               (&input)[ITEMS_PER_THREAD],         ///< [in] Calling thread's input items
        FlagOp          flag_op)                            ///< [in] Binary boolean flag predicate
    {
        T preds[ITEMS_PER_THREAD];
        FlagHeads(head_flags, input, preds, flag_op);
    }


    template <
        int             ITEMS_PER_THREAD,
        typename        FlagT,
        typename        FlagOp>
    __device__ __forceinline__ void FlagHeads(
        FlagT           (&head_flags)[ITEMS_PER_THREAD],    ///< [out] Calling thread's discontinuity head_flags
        T               (&input)[ITEMS_PER_THREAD],         ///< [in] Calling thread's input items
        FlagOp          flag_op,                            ///< [in] Binary boolean flag predicate
        T               tile_predecessor_item)              ///< [in] <b>[<em>thread</em><sub>0</sub> only]</b> Item with which to compare the first tile item (<tt>input<sub>0</sub></tt> from <em>thread</em><sub>0</sub>).
    {
        T preds[ITEMS_PER_THREAD];
        FlagHeads(head_flags, input, preds, flag_op, tile_predecessor_item);
    }



    template <
        int             ITEMS_PER_THREAD,
        typename        FlagT,
        typename        FlagOp>
    __device__ __forceinline__ void FlagTails(
        FlagT           (&tail_flags)[ITEMS_PER_THREAD],    ///< [out] Calling thread's discontinuity tail_flags
        T               (&input)[ITEMS_PER_THREAD],         ///< [in] Calling thread's input items
        FlagOp          flag_op)                            ///< [in] Binary boolean flag predicate
    {
        // Share first item
        temp_storage.first_items[linear_tid] = input[0];

        CTA_SYNC();

        // Set flag for last thread-item
        tail_flags[ITEMS_PER_THREAD - 1] = (linear_tid == BLOCK_THREADS - 1) ?
            1 :                             // Last thread
            ApplyOp<FlagOp>::FlagT(
                flag_op,
                input[ITEMS_PER_THREAD - 1],
                temp_storage.first_items[linear_tid + 1],
                (linear_tid * ITEMS_PER_THREAD) + ITEMS_PER_THREAD);

        // Set tail_flags for remaining items
        Iterate<0, ITEMS_PER_THREAD - 1>::FlagTails(linear_tid, tail_flags, input, flag_op);
    }


    template <
        int             ITEMS_PER_THREAD,
        typename        FlagT,
        typename        FlagOp>
    __device__ __forceinline__ void FlagTails(
        FlagT           (&tail_flags)[ITEMS_PER_THREAD],    ///< [out] Calling thread's discontinuity tail_flags
        T               (&input)[ITEMS_PER_THREAD],         ///< [in] Calling thread's input items
        FlagOp          flag_op,                            ///< [in] Binary boolean flag predicate
        T               tile_successor_item)                ///< [in] <b>[<em>thread</em><sub><tt>BLOCK_THREADS</tt>-1</sub> only]</b> Item with which to compare the last tile item (<tt>input</tt><sub><em>ITEMS_PER_THREAD</em>-1</sub> from <em>thread</em><sub><em>BLOCK_THREADS</em>-1</sub>).
    {
        // Share first item
        temp_storage.first_items[linear_tid] = input[0];

        CTA_SYNC();

        // Set flag for last thread-item
        T successor_item = (linear_tid == BLOCK_THREADS - 1) ?
            tile_successor_item :              // Last thread
            temp_storage.first_items[linear_tid + 1];

        tail_flags[ITEMS_PER_THREAD - 1] = ApplyOp<FlagOp>::FlagT(
            flag_op,
            input[ITEMS_PER_THREAD - 1],
            successor_item,
            (linear_tid * ITEMS_PER_THREAD) + ITEMS_PER_THREAD);

        // Set tail_flags for remaining items
        Iterate<0, ITEMS_PER_THREAD - 1>::FlagTails(linear_tid, tail_flags, input, flag_op);
    }


    template <
        int             ITEMS_PER_THREAD,
        typename        FlagT,
        typename        FlagOp>
    __device__ __forceinline__ void FlagHeadsAndTails(
        FlagT           (&head_flags)[ITEMS_PER_THREAD],    ///< [out] Calling thread's discontinuity head_flags
        FlagT           (&tail_flags)[ITEMS_PER_THREAD],    ///< [out] Calling thread's discontinuity tail_flags
        T               (&input)[ITEMS_PER_THREAD],         ///< [in] Calling thread's input items
        FlagOp          flag_op)                            ///< [in] Binary boolean flag predicate
    {
        // Share first and last items
        temp_storage.first_items[linear_tid] = input[0];
        temp_storage.last_items[linear_tid] = input[ITEMS_PER_THREAD - 1];

        CTA_SYNC();

        T preds[ITEMS_PER_THREAD];

        // Set flag for first thread-item
        preds[0] = temp_storage.last_items[linear_tid - 1];
        if (linear_tid == 0)
        {
            head_flags[0] = 1;
        }
        else
        {
            head_flags[0] = ApplyOp<FlagOp>::FlagT(
                flag_op,
                preds[0],
                input[0],
                linear_tid * ITEMS_PER_THREAD);
        }


        // Set flag for last thread-item
        tail_flags[ITEMS_PER_THREAD - 1] = (linear_tid == BLOCK_THREADS - 1) ?
            1 :                             // Last thread
            ApplyOp<FlagOp>::FlagT(
                flag_op,
                input[ITEMS_PER_THREAD - 1],
                temp_storage.first_items[linear_tid + 1],
                (linear_tid * ITEMS_PER_THREAD) + ITEMS_PER_THREAD);

        // Set head_flags for remaining items
        Iterate<1, ITEMS_PER_THREAD>::FlagHeads(linear_tid, head_flags, input, preds, flag_op);

        // Set tail_flags for remaining items
        Iterate<0, ITEMS_PER_THREAD - 1>::FlagTails(linear_tid, tail_flags, input, flag_op);
    }


    template <
        int             ITEMS_PER_THREAD,
        typename        FlagT,
        typename        FlagOp>
    __device__ __forceinline__ void FlagHeadsAndTails(
        FlagT           (&head_flags)[ITEMS_PER_THREAD],    ///< [out] Calling thread's discontinuity head_flags
        FlagT           (&tail_flags)[ITEMS_PER_THREAD],    ///< [out] Calling thread's discontinuity tail_flags
        T               tile_successor_item,                ///< [in] <b>[<em>thread</em><sub><tt>BLOCK_THREADS</tt>-1</sub> only]</b> Item with which to compare the last tile item (<tt>input</tt><sub><em>ITEMS_PER_THREAD</em>-1</sub> from <em>thread</em><sub><em>BLOCK_THREADS</em>-1</sub>).
        T               (&input)[ITEMS_PER_THREAD],         ///< [in] Calling thread's input items
        FlagOp          flag_op)                            ///< [in] Binary boolean flag predicate
    {
        // Share first and last items
        temp_storage.first_items[linear_tid] = input[0];
        temp_storage.last_items[linear_tid] = input[ITEMS_PER_THREAD - 1];

        CTA_SYNC();

        T preds[ITEMS_PER_THREAD];

        // Set flag for first thread-item
        if (linear_tid == 0)
        {
            head_flags[0] = 1;
        }
        else
        {
            preds[0] = temp_storage.last_items[linear_tid - 1];
            head_flags[0] = ApplyOp<FlagOp>::FlagT(
                flag_op,
                preds[0],
                input[0],
                linear_tid * ITEMS_PER_THREAD);
        }

        // Set flag for last thread-item
        T successor_item = (linear_tid == BLOCK_THREADS - 1) ?
            tile_successor_item :              // Last thread
            temp_storage.first_items[linear_tid + 1];

        tail_flags[ITEMS_PER_THREAD - 1] = ApplyOp<FlagOp>::FlagT(
            flag_op,
            input[ITEMS_PER_THREAD - 1],
            successor_item,
            (linear_tid * ITEMS_PER_THREAD) + ITEMS_PER_THREAD);

        // Set head_flags for remaining items
        Iterate<1, ITEMS_PER_THREAD>::FlagHeads(linear_tid, head_flags, input, preds, flag_op);

        // Set tail_flags for remaining items
        Iterate<0, ITEMS_PER_THREAD - 1>::FlagTails(linear_tid, tail_flags, input, flag_op);
    }

    template <
        int             ITEMS_PER_THREAD,
        typename        FlagT,
        typename        FlagOp>
    __device__ __forceinline__ void FlagHeadsAndTails(
        FlagT           (&head_flags)[ITEMS_PER_THREAD],    ///< [out] Calling thread's discontinuity head_flags
        T               tile_predecessor_item,              ///< [in] <b>[<em>thread</em><sub>0</sub> only]</b> Item with which to compare the first tile item (<tt>input<sub>0</sub></tt> from <em>thread</em><sub>0</sub>).
        FlagT           (&tail_flags)[ITEMS_PER_THREAD],    ///< [out] Calling thread's discontinuity tail_flags
        T               (&input)[ITEMS_PER_THREAD],         ///< [in] Calling thread's input items
        FlagOp          flag_op)                            ///< [in] Binary boolean flag predicate
    {
        // Share first and last items
        temp_storage.first_items[linear_tid] = input[0];
        temp_storage.last_items[linear_tid] = input[ITEMS_PER_THREAD - 1];

        CTA_SYNC();

        T preds[ITEMS_PER_THREAD];

        // Set flag for first thread-item
        preds[0] = (linear_tid == 0) ?
            tile_predecessor_item :              // First thread
            temp_storage.last_items[linear_tid - 1];

        head_flags[0] = ApplyOp<FlagOp>::FlagT(
            flag_op,
            preds[0],
            input[0],
            linear_tid * ITEMS_PER_THREAD);

        // Set flag for last thread-item
        tail_flags[ITEMS_PER_THREAD - 1] = (linear_tid == BLOCK_THREADS - 1) ?
            1 :                             // Last thread
            ApplyOp<FlagOp>::FlagT(
                flag_op,
                input[ITEMS_PER_THREAD - 1],
                temp_storage.first_items[linear_tid + 1],
                (linear_tid * ITEMS_PER_THREAD) + ITEMS_PER_THREAD);

        // Set head_flags for remaining items
        Iterate<1, ITEMS_PER_THREAD>::FlagHeads(linear_tid, head_flags, input, preds, flag_op);

        // Set tail_flags for remaining items
        Iterate<0, ITEMS_PER_THREAD - 1>::FlagTails(linear_tid, tail_flags, input, flag_op);
    }


    template <
        int             ITEMS_PER_THREAD,
        typename        FlagT,
        typename        FlagOp>
    __device__ __forceinline__ void FlagHeadsAndTails(
        FlagT           (&head_flags)[ITEMS_PER_THREAD],    ///< [out] Calling thread's discontinuity head_flags
        T               tile_predecessor_item,              ///< [in] <b>[<em>thread</em><sub>0</sub> only]</b> Item with which to compare the first tile item (<tt>input<sub>0</sub></tt> from <em>thread</em><sub>0</sub>).
        FlagT           (&tail_flags)[ITEMS_PER_THREAD],    ///< [out] Calling thread's discontinuity tail_flags
        T               tile_successor_item,                ///< [in] <b>[<em>thread</em><sub><tt>BLOCK_THREADS</tt>-1</sub> only]</b> Item with which to compare the last tile item (<tt>input</tt><sub><em>ITEMS_PER_THREAD</em>-1</sub> from <em>thread</em><sub><em>BLOCK_THREADS</em>-1</sub>).
        T               (&input)[ITEMS_PER_THREAD],         ///< [in] Calling thread's input items
        FlagOp          flag_op)                            ///< [in] Binary boolean flag predicate
    {
        // Share first and last items
        temp_storage.first_items[linear_tid] = input[0];
        temp_storage.last_items[linear_tid] = input[ITEMS_PER_THREAD - 1];

        CTA_SYNC();

        T preds[ITEMS_PER_THREAD];

        // Set flag for first thread-item
        preds[0] = (linear_tid == 0) ?
            tile_predecessor_item :              // First thread
            temp_storage.last_items[linear_tid - 1];

        head_flags[0] = ApplyOp<FlagOp>::FlagT(
            flag_op,
            preds[0],
            input[0],
            linear_tid * ITEMS_PER_THREAD);

        // Set flag for last thread-item
        T successor_item = (linear_tid == BLOCK_THREADS - 1) ?
            tile_successor_item :              // Last thread
            temp_storage.first_items[linear_tid + 1];

        tail_flags[ITEMS_PER_THREAD - 1] = ApplyOp<FlagOp>::FlagT(
            flag_op,
            input[ITEMS_PER_THREAD - 1],
            successor_item,
            (linear_tid * ITEMS_PER_THREAD) + ITEMS_PER_THREAD);

        // Set head_flags for remaining items
        Iterate<1, ITEMS_PER_THREAD>::FlagHeads(linear_tid, head_flags, input, preds, flag_op);

        // Set tail_flags for remaining items
        Iterate<0, ITEMS_PER_THREAD - 1>::FlagTails(linear_tid, tail_flags, input, flag_op);
    }



};


}               // CUB namespace
CUB_NS_POSTFIX  // Optional outer namespace(s)