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
|
/******************************************************************************
* Copyright (c) 2016, 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.
*
******************************************************************************/
#pragma once
#include <thrust/detail/config.h>
#if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
#include <thrust/detail/cstdint.h>
#include <thrust/detail/temporary_array.h>
#include <thrust/system/cuda/detail/util.h>
#include <thrust/system/cuda/config.h>
#include <thrust/system/cuda/detail/core/agent_launcher.h>
#include <thrust/system/cuda/detail/core/util.h>
#include <cub/device/device_radix_sort.cuh>
#include <cub/device/device_merge_sort.cuh>
#include <thrust/system/cuda/detail/execution_policy.h>
#include <thrust/system/cuda/detail/par_to_seq.h>
#include <thrust/detail/trivial_sequence.h>
#include <thrust/detail/integer_math.h>
#include <thrust/extrema.h>
#include <thrust/sort.h>
#include <thrust/distance.h>
#include <thrust/sequence.h>
#include <thrust/detail/alignment.h>
#include <thrust/type_traits/is_contiguous_iterator.h>
THRUST_NAMESPACE_BEGIN
namespace cuda_cub {
namespace __merge_sort {
template <class KeysIt,
class ItemsIt,
class Size,
class CompareOp>
THRUST_RUNTIME_FUNCTION cudaError_t
doit_step(void* d_temp_storage,
size_t& temp_storage_bytes,
KeysIt keys,
ItemsIt ,
Size keys_count,
CompareOp compare_op,
cudaStream_t stream,
bool debug_sync,
thrust::detail::integral_constant<bool, false> /* sort_keys */)
{
using ItemsInputIt = cub::NullType *;
ItemsInputIt items = nullptr;
using DispatchMergeSortT = cub::DispatchMergeSort<KeysIt,
ItemsInputIt,
KeysIt,
ItemsInputIt,
Size,
CompareOp>;
return DispatchMergeSortT::Dispatch(d_temp_storage,
temp_storage_bytes,
keys,
items,
keys,
items,
keys_count,
compare_op,
stream,
debug_sync);
}
template <class KeysIt,
class ItemsIt,
class Size,
class CompareOp>
THRUST_RUNTIME_FUNCTION cudaError_t
doit_step(void *d_temp_storage,
size_t &temp_storage_bytes,
KeysIt keys,
ItemsIt items,
Size keys_count,
CompareOp compare_op,
cudaStream_t stream,
bool debug_sync,
thrust::detail::integral_constant<bool, true> /* sort_items */)
{
using DispatchMergeSortT =
cub::DispatchMergeSort<KeysIt, ItemsIt, KeysIt, ItemsIt, Size, CompareOp>;
return DispatchMergeSortT::Dispatch(d_temp_storage,
temp_storage_bytes,
keys,
items,
keys,
items,
keys_count,
compare_op,
stream,
debug_sync);
}
template <class SORT_ITEMS,
class /* STABLE */,
class KeysIt,
class ItemsIt,
class Size,
class CompareOp>
THRUST_RUNTIME_FUNCTION cudaError_t
doit_step(void *d_temp_storage,
size_t &temp_storage_bytes,
KeysIt keys,
ItemsIt items,
Size keys_count,
CompareOp compare_op,
cudaStream_t stream,
bool debug_sync)
{
if (keys_count == 0)
{
return cudaSuccess;
}
thrust::detail::integral_constant<bool, SORT_ITEMS::value> sort_items{};
return doit_step(d_temp_storage,
temp_storage_bytes,
keys,
items,
keys_count,
compare_op,
stream,
debug_sync,
sort_items);
}
template <typename SORT_ITEMS,
typename STABLE,
typename Derived,
typename KeysIt,
typename ItemsIt,
typename CompareOp>
THRUST_RUNTIME_FUNCTION
void merge_sort(execution_policy<Derived>& policy,
KeysIt keys_first,
KeysIt keys_last,
ItemsIt items_first,
CompareOp compare_op)
{
typedef typename iterator_traits<KeysIt>::difference_type size_type;
size_type count = static_cast<size_type>(thrust::distance(keys_first, keys_last));
size_t storage_size = 0;
cudaStream_t stream = cuda_cub::stream(policy);
bool debug_sync = THRUST_DEBUG_SYNC_FLAG;
cudaError_t status;
status = doit_step<SORT_ITEMS, STABLE>(NULL,
storage_size,
keys_first,
items_first,
count,
compare_op,
stream,
debug_sync);
cuda_cub::throw_on_error(status, "merge_sort: failed on 1st step");
// Allocate temporary storage.
thrust::detail::temporary_array<thrust::detail::uint8_t, Derived>
tmp(policy, storage_size);
void *ptr = static_cast<void*>(tmp.data().get());
status = doit_step<SORT_ITEMS, STABLE>(ptr,
storage_size,
keys_first,
items_first,
count,
compare_op,
stream,
debug_sync);
cuda_cub::throw_on_error(status, "merge_sort: failed on 2nd step");
status = cuda_cub::synchronize_optional(policy);
cuda_cub::throw_on_error(status, "merge_sort: failed to synchronize");
}
} // namespace __merge_sort
namespace __radix_sort {
template <class SORT_ITEMS, class Comparator>
struct dispatch;
// sort keys in ascending order
template <class K>
struct dispatch<thrust::detail::false_type, thrust::less<K> >
{
template <class Key, class Item, class Size>
THRUST_RUNTIME_FUNCTION static cudaError_t
doit(void* d_temp_storage,
size_t& temp_storage_bytes,
cub::DoubleBuffer<Key>& keys_buffer,
cub::DoubleBuffer<Item>& /*items_buffer*/,
Size count,
cudaStream_t stream,
bool debug_sync)
{
return cub::DeviceRadixSort::SortKeys(d_temp_storage,
temp_storage_bytes,
keys_buffer,
static_cast<int>(count),
0,
static_cast<int>(sizeof(Key) * 8),
stream,
debug_sync);
}
}; // struct dispatch -- sort keys in ascending order;
// sort keys in descending order
template <class K>
struct dispatch<thrust::detail::false_type, thrust::greater<K> >
{
template <class Key, class Item, class Size>
THRUST_RUNTIME_FUNCTION static cudaError_t
doit(void* d_temp_storage,
size_t& temp_storage_bytes,
cub::DoubleBuffer<Key>& keys_buffer,
cub::DoubleBuffer<Item>& /*items_buffer*/,
Size count,
cudaStream_t stream,
bool debug_sync)
{
return cub::DeviceRadixSort::SortKeysDescending(d_temp_storage,
temp_storage_bytes,
keys_buffer,
static_cast<int>(count),
0,
static_cast<int>(sizeof(Key) * 8),
stream,
debug_sync);
}
}; // struct dispatch -- sort keys in descending order;
// sort pairs in ascending order
template <class K>
struct dispatch<thrust::detail::true_type, thrust::less<K> >
{
template <class Key, class Item, class Size>
THRUST_RUNTIME_FUNCTION static cudaError_t
doit(void* d_temp_storage,
size_t& temp_storage_bytes,
cub::DoubleBuffer<Key>& keys_buffer,
cub::DoubleBuffer<Item>& items_buffer,
Size count,
cudaStream_t stream,
bool debug_sync)
{
return cub::DeviceRadixSort::SortPairs(d_temp_storage,
temp_storage_bytes,
keys_buffer,
items_buffer,
static_cast<int>(count),
0,
static_cast<int>(sizeof(Key) * 8),
stream,
debug_sync);
}
}; // struct dispatch -- sort pairs in ascending order;
// sort pairs in descending order
template <class K>
struct dispatch<thrust::detail::true_type, thrust::greater<K> >
{
template <class Key, class Item, class Size>
THRUST_RUNTIME_FUNCTION static cudaError_t
doit(void* d_temp_storage,
size_t& temp_storage_bytes,
cub::DoubleBuffer<Key>& keys_buffer,
cub::DoubleBuffer<Item>& items_buffer,
Size count,
cudaStream_t stream,
bool debug_sync)
{
return cub::DeviceRadixSort::SortPairsDescending(d_temp_storage,
temp_storage_bytes,
keys_buffer,
items_buffer,
static_cast<int>(count),
0,
static_cast<int>(sizeof(Key) * 8),
stream,
debug_sync);
}
}; // struct dispatch -- sort pairs in descending order;
template <typename SORT_ITEMS,
typename Derived,
typename Key,
typename Item,
typename Size,
typename CompareOp>
THRUST_RUNTIME_FUNCTION
void radix_sort(execution_policy<Derived>& policy,
Key* keys,
Item* items,
Size count,
CompareOp)
{
size_t temp_storage_bytes = 0;
cudaStream_t stream = cuda_cub::stream(policy);
bool debug_sync = THRUST_DEBUG_SYNC_FLAG;
cub::DoubleBuffer<Key> keys_buffer(keys, NULL);
cub::DoubleBuffer<Item> items_buffer(items, NULL);
Size keys_count = count;
Size items_count = SORT_ITEMS::value ? count : 0;
cudaError_t status;
status = dispatch<SORT_ITEMS, CompareOp>::doit(NULL,
temp_storage_bytes,
keys_buffer,
items_buffer,
keys_count,
stream,
debug_sync);
cuda_cub::throw_on_error(status, "radix_sort: failed on 1st step");
size_t keys_temp_storage = core::align_to(sizeof(Key) * keys_count, 128);
size_t items_temp_storage = core::align_to(sizeof(Item) * items_count, 128);
size_t storage_size = keys_temp_storage
+ items_temp_storage
+ temp_storage_bytes;
// Allocate temporary storage.
thrust::detail::temporary_array<thrust::detail::uint8_t, Derived>
tmp(policy, storage_size);
keys_buffer.d_buffers[1] = thrust::detail::aligned_reinterpret_cast<Key*>(
tmp.data().get()
);
items_buffer.d_buffers[1] = thrust::detail::aligned_reinterpret_cast<Item*>(
tmp.data().get() + keys_temp_storage
);
void *ptr = static_cast<void*>(
tmp.data().get() + keys_temp_storage + items_temp_storage
);
status = dispatch<SORT_ITEMS, CompareOp>::doit(ptr,
temp_storage_bytes,
keys_buffer,
items_buffer,
keys_count,
stream,
debug_sync);
cuda_cub::throw_on_error(status, "radix_sort: failed on 2nd step");
if (keys_buffer.selector != 0)
{
Key* temp_ptr = reinterpret_cast<Key*>(keys_buffer.d_buffers[1]);
cuda_cub::copy_n(policy, temp_ptr, keys_count, keys);
}
THRUST_IF_CONSTEXPR(SORT_ITEMS::value)
{
if (items_buffer.selector != 0)
{
Item *temp_ptr = reinterpret_cast<Item *>(items_buffer.d_buffers[1]);
cuda_cub::copy_n(policy, temp_ptr, items_count, items);
}
}
}
} // __radix_sort
//---------------------------------------------------------------------
// Smart sort picks at compile-time whether to dispatch radix or merge sort
//---------------------------------------------------------------------
namespace __smart_sort {
template <class Key, class CompareOp>
struct can_use_primitive_sort
: thrust::detail::and_<
thrust::detail::is_arithmetic<Key>,
thrust::detail::or_<
thrust::detail::is_same<CompareOp, thrust::less<Key> >,
thrust::detail::is_same<CompareOp, thrust::greater<Key> > > > {};
template <class Iterator, class CompareOp>
struct enable_if_primitive_sort
: thrust::detail::enable_if<
can_use_primitive_sort<typename iterator_value<Iterator>::type,
CompareOp>::value> {};
template <class Iterator, class CompareOp>
struct enable_if_comparison_sort
: thrust::detail::disable_if<
can_use_primitive_sort<typename iterator_value<Iterator>::type,
CompareOp>::value> {};
template <class SORT_ITEMS,
class STABLE,
class Policy,
class KeysIt,
class ItemsIt,
class CompareOp>
THRUST_RUNTIME_FUNCTION typename enable_if_comparison_sort<KeysIt, CompareOp>::type
smart_sort(Policy& policy,
KeysIt keys_first,
KeysIt keys_last,
ItemsIt items_first,
CompareOp compare_op)
{
__merge_sort::merge_sort<SORT_ITEMS, STABLE>(policy,
keys_first,
keys_last,
items_first,
compare_op);
}
template <class SORT_ITEMS,
class STABLE,
class Policy,
class KeysIt,
class ItemsIt,
class CompareOp>
THRUST_RUNTIME_FUNCTION typename enable_if_primitive_sort<KeysIt, CompareOp>::type
smart_sort(execution_policy<Policy>& policy,
KeysIt keys_first,
KeysIt keys_last,
ItemsIt items_first,
CompareOp compare_op)
{
// ensure sequences have trivial iterators
thrust::detail::trivial_sequence<KeysIt, Policy>
keys(policy, keys_first, keys_last);
if (SORT_ITEMS::value)
{
thrust::detail::trivial_sequence<ItemsIt, Policy>
values(policy, items_first, items_first + (keys_last - keys_first));
__radix_sort::radix_sort<SORT_ITEMS>(
policy,
thrust::raw_pointer_cast(&*keys.begin()),
thrust::raw_pointer_cast(&*values.begin()),
keys_last - keys_first,
compare_op);
if (!is_contiguous_iterator<ItemsIt>::value)
{
cuda_cub::copy(policy, values.begin(), values.end(), items_first);
}
}
else
{
__radix_sort::radix_sort<SORT_ITEMS>(
policy,
thrust::raw_pointer_cast(&*keys.begin()),
thrust::raw_pointer_cast(&*keys.begin()),
keys_last - keys_first,
compare_op);
}
// copy results back, if necessary
if (!is_contiguous_iterator<KeysIt>::value)
{
cuda_cub::copy(policy, keys.begin(), keys.end(), keys_first);
}
cuda_cub::throw_on_error(
cuda_cub::synchronize_optional(policy),
"smart_sort: failed to synchronize");
}
} // namespace __smart_sort
//-------------------------
// Thrust API entry points
//-------------------------
__thrust_exec_check_disable__
template <class Derived, class ItemsIt, class CompareOp>
void __host__ __device__
sort(execution_policy<Derived>& policy,
ItemsIt first,
ItemsIt last,
CompareOp compare_op)
{
if (__THRUST_HAS_CUDART__)
{
typedef typename thrust::iterator_value<ItemsIt>::type item_type;
__smart_sort::smart_sort<thrust::detail::false_type, thrust::detail::false_type>(
policy, first, last, (item_type*)NULL, compare_op);
}
else
{
#if !__THRUST_HAS_CUDART__
thrust::sort(cvt_to_seq(derived_cast(policy)), first, last, compare_op);
#endif
}
}
__thrust_exec_check_disable__
template <class Derived, class ItemsIt, class CompareOp>
void __host__ __device__
stable_sort(execution_policy<Derived>& policy,
ItemsIt first,
ItemsIt last,
CompareOp compare_op)
{
if (__THRUST_HAS_CUDART__)
{
typedef typename thrust::iterator_value<ItemsIt>::type item_type;
__smart_sort::smart_sort<thrust::detail::false_type, thrust::detail::true_type>(
policy, first, last, (item_type*)NULL, compare_op);
}
else
{
#if !__THRUST_HAS_CUDART__
thrust::stable_sort(cvt_to_seq(derived_cast(policy)), first, last, compare_op);
#endif
}
}
__thrust_exec_check_disable__
template <class Derived, class KeysIt, class ValuesIt, class CompareOp>
void __host__ __device__
sort_by_key(execution_policy<Derived>& policy,
KeysIt keys_first,
KeysIt keys_last,
ValuesIt values,
CompareOp compare_op)
{
if (__THRUST_HAS_CUDART__)
{
__smart_sort::smart_sort<thrust::detail::true_type, thrust::detail::false_type>(
policy, keys_first, keys_last, values, compare_op);
}
else
{
#if !__THRUST_HAS_CUDART__
thrust::sort_by_key(
cvt_to_seq(derived_cast(policy)), keys_first, keys_last, values, compare_op);
#endif
}
}
__thrust_exec_check_disable__
template <class Derived,
class KeysIt,
class ValuesIt,
class CompareOp>
void __host__ __device__
stable_sort_by_key(execution_policy<Derived> &policy,
KeysIt keys_first,
KeysIt keys_last,
ValuesIt values,
CompareOp compare_op)
{
if (__THRUST_HAS_CUDART__)
{
__smart_sort::smart_sort<thrust::detail::true_type, thrust::detail::true_type>(
policy, keys_first, keys_last, values, compare_op);
}
else
{
#if !__THRUST_HAS_CUDART__
thrust::stable_sort_by_key(
cvt_to_seq(derived_cast(policy)), keys_first, keys_last, values, compare_op);
#endif
}
}
// API with default comparator
template <class Derived, class ItemsIt>
void __host__ __device__
sort(execution_policy<Derived>& policy,
ItemsIt first,
ItemsIt last)
{
typedef typename thrust::iterator_value<ItemsIt>::type item_type;
cuda_cub::sort(policy, first, last, less<item_type>());
}
template <class Derived, class ItemsIt>
void __host__ __device__
stable_sort(execution_policy<Derived>& policy,
ItemsIt first,
ItemsIt last)
{
typedef typename thrust::iterator_value<ItemsIt>::type item_type;
cuda_cub::stable_sort(policy, first, last, less<item_type>());
}
template <class Derived, class KeysIt, class ValuesIt>
void __host__ __device__
sort_by_key(execution_policy<Derived>& policy,
KeysIt keys_first,
KeysIt keys_last,
ValuesIt values)
{
typedef typename thrust::iterator_value<KeysIt>::type key_type;
cuda_cub::sort_by_key(policy, keys_first, keys_last, values, less<key_type>());
}
template <class Derived, class KeysIt, class ValuesIt>
void __host__ __device__
stable_sort_by_key(
execution_policy<Derived>& policy, KeysIt keys_first, KeysIt keys_last, ValuesIt values)
{
typedef typename thrust::iterator_value<KeysIt>::type key_type;
cuda_cub::stable_sort_by_key(policy, keys_first, keys_last, values, less<key_type>());
}
} // namespace cuda_cub
THRUST_NAMESPACE_END
#endif
|
