Blame view
tools/cub-1.8.0/cub/device/device_segmented_reduce.cuh
35.7 KB
8dcb6dfcb
first commit
|
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 |
/******************************************************************************
* 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
* cub::DeviceSegmentedReduce provides device-wide, parallel operations for computing a batched reduction across multiple sequences of data items residing within device-accessible memory.
*/
#pragma once
#include <stdio.h>
#include <iterator>
#include "../iterator/arg_index_input_iterator.cuh"
#include "dispatch/dispatch_reduce.cuh"
#include "dispatch/dispatch_reduce_by_key.cuh"
#include "../util_type.cuh"
#include "../util_namespace.cuh"
/// Optional outer namespace(s)
CUB_NS_PREFIX
/// CUB namespace
namespace cub {
/**
* \brief DeviceSegmentedReduce provides device-wide, parallel operations for computing a reduction across multiple sequences of data items residing within device-accessible memory. 
* \ingroup SegmentedModule
*
* \par Overview
* A <a href="http://en.wikipedia.org/wiki/Reduce_(higher-order_function)"><em>reduction</em></a> (or <em>fold</em>)
* uses a binary combining operator to compute a single aggregate from a sequence of input elements.
*
* \par Usage Considerations
* \cdp_class{DeviceSegmentedReduce}
*
*/
struct DeviceSegmentedReduce
{
/**
* \brief Computes a device-wide segmented reduction using the specified binary \p reduction_op functor.
*
* \par
* - Does not support binary reduction operators that are non-commutative.
* - When input a contiguous sequence of segments, a single sequence
* \p segment_offsets (of length <tt>num_segments+1</tt>) can be aliased
* for both the \p d_begin_offsets and \p d_end_offsets parameters (where
* the latter is specified as <tt>segment_offsets+1</tt>).
* - \devicestorage
*
* \par Snippet
* The code snippet below illustrates a custom min-reduction of a device vector of \p int data elements.
* \par
* \code
* #include <cub/cub.cuh> // or equivalently <cub/device/device_radix_sort.cuh>
*
* // CustomMin functor
* struct CustomMin
* {
* template <typename T>
* CUB_RUNTIME_FUNCTION __forceinline__
* T operator()(const T &a, const T &b) const {
* return (b < a) ? b : a;
* }
* };
*
* // Declare, allocate, and initialize device-accessible pointers for input and output
* int num_segments; // e.g., 3
* int *d_offsets; // e.g., [0, 3, 3, 7]
* int *d_in; // e.g., [8, 6, 7, 5, 3, 0, 9]
* int *d_out; // e.g., [-, -, -]
* CustomMin min_op;
* int initial_value; // e.g., INT_MAX
* ...
*
* // Determine temporary device storage requirements
* void *d_temp_storage = NULL;
* size_t temp_storage_bytes = 0;
* cub::DeviceSegmentedReduce::Reduce(d_temp_storage, temp_storage_bytes, d_in, d_out,
* num_segments, d_offsets, d_offsets + 1, min_op, initial_value);
*
* // Allocate temporary storage
* cudaMalloc(&d_temp_storage, temp_storage_bytes);
*
* // Run reduction
* cub::DeviceSegmentedReduce::Reduce(d_temp_storage, temp_storage_bytes, d_in, d_out,
* num_segments, d_offsets, d_offsets + 1, min_op, initial_value);
*
* // d_out <-- [6, INT_MAX, 0]
*
* \endcode
*
* \tparam InputIteratorT <b>[inferred]</b> Random-access input iterator type for reading input items \iterator
* \tparam OutputIteratorT <b>[inferred]</b> Output iterator type for recording the reduced aggregate \iterator
* \tparam OffsetIteratorT <b>[inferred]</b> Random-access input iterator type for reading segment offsets \iterator
* \tparam ReductionOp <b>[inferred]</b> Binary reduction functor type having member <tt>T operator()(const T &a, const T &b)</tt>
* \tparam T <b>[inferred]</b> Data element type that is convertible to the \p value type of \p InputIteratorT
*/
template <
typename InputIteratorT,
typename OutputIteratorT,
typename OffsetIteratorT,
typename ReductionOp,
typename T>
CUB_RUNTIME_FUNCTION
static cudaError_t Reduce(
void *d_temp_storage, ///< [in] %Device-accessible allocation of temporary storage. When NULL, the required allocation size is written to \p temp_storage_bytes and no work is done.
size_t &temp_storage_bytes, ///< [in,out] Reference to size in bytes of \p d_temp_storage allocation
InputIteratorT d_in, ///< [in] Pointer to the input sequence of data items
OutputIteratorT d_out, ///< [out] Pointer to the output aggregate
int num_segments, ///< [in] The number of segments that comprise the sorting data
OffsetIteratorT d_begin_offsets, ///< [in] Pointer to the sequence of beginning offsets of length \p num_segments, such that <tt>d_begin_offsets[i]</tt> is the first element of the <em>i</em><sup>th</sup> data segment in <tt>d_keys_*</tt> and <tt>d_values_*</tt>
OffsetIteratorT d_end_offsets, ///< [in] Pointer to the sequence of ending offsets of length \p num_segments, such that <tt>d_end_offsets[i]-1</tt> is the last element of the <em>i</em><sup>th</sup> data segment in <tt>d_keys_*</tt> and <tt>d_values_*</tt>. If <tt>d_end_offsets[i]-1</tt> <= <tt>d_begin_offsets[i]</tt>, the <em>i</em><sup>th</sup> is considered empty.
ReductionOp reduction_op, ///< [in] Binary reduction functor
T initial_value, ///< [in] Initial value of the reduction for each segment
cudaStream_t stream = 0, ///< [in] <b>[optional]</b> CUDA stream to launch kernels within. Default is stream<sub>0</sub>.
bool debug_synchronous = false) ///< [in] <b>[optional]</b> Whether or not to synchronize the stream after every kernel launch to check for errors. Also causes launch configurations to be printed to the console. Default is \p false.
{
// Signed integer type for global offsets
typedef int OffsetT;
return DispatchSegmentedReduce<InputIteratorT, OutputIteratorT, OffsetIteratorT, OffsetT, ReductionOp>::Dispatch(
d_temp_storage,
temp_storage_bytes,
d_in,
d_out,
num_segments,
d_begin_offsets,
d_end_offsets,
reduction_op,
initial_value,
stream,
debug_synchronous);
}
/**
* \brief Computes a device-wide segmented sum using the addition ('+') operator.
*
* \par
* - Uses \p 0 as the initial value of the reduction for each segment.
* - When input a contiguous sequence of segments, a single sequence
* \p segment_offsets (of length <tt>num_segments+1</tt>) can be aliased
* for both the \p d_begin_offsets and \p d_end_offsets parameters (where
* the latter is specified as <tt>segment_offsets+1</tt>).
* - Does not support \p + operators that are non-commutative..
* - \devicestorage
*
* \par Snippet
* The code snippet below illustrates the sum reduction of a device vector of \p int data elements.
* \par
* \code
* #include <cub/cub.cuh> // or equivalently <cub/device/device_radix_sort.cuh>
*
* // Declare, allocate, and initialize device-accessible pointers for input and output
* int num_segments; // e.g., 3
* int *d_offsets; // e.g., [0, 3, 3, 7]
* int *d_in; // e.g., [8, 6, 7, 5, 3, 0, 9]
* int *d_out; // e.g., [-, -, -]
* ...
*
* // Determine temporary device storage requirements
* void *d_temp_storage = NULL;
* size_t temp_storage_bytes = 0;
* cub::DeviceSegmentedReduce::Sum(d_temp_storage, temp_storage_bytes, d_in, d_out,
* num_segments, d_offsets, d_offsets + 1);
*
* // Allocate temporary storage
* cudaMalloc(&d_temp_storage, temp_storage_bytes);
*
* // Run sum-reduction
* cub::DeviceSegmentedReduce::Sum(d_temp_storage, temp_storage_bytes, d_in, d_out,
* num_segments, d_offsets, d_offsets + 1);
*
* // d_out <-- [21, 0, 17]
*
* \endcode
*
* \tparam InputIteratorT <b>[inferred]</b> Random-access input iterator type for reading input items \iterator
* \tparam OutputIteratorT <b>[inferred]</b> Output iterator type for recording the reduced aggregate \iterator
* \tparam OffsetIteratorT <b>[inferred]</b> Random-access input iterator type for reading segment offsets \iterator
*/
template <
typename InputIteratorT,
typename OutputIteratorT,
typename OffsetIteratorT>
CUB_RUNTIME_FUNCTION
static cudaError_t Sum(
void *d_temp_storage, ///< [in] %Device-accessible allocation of temporary storage. When NULL, the required allocation size is written to \p temp_storage_bytes and no work is done.
size_t &temp_storage_bytes, ///< [in,out] Reference to size in bytes of \p d_temp_storage allocation
InputIteratorT d_in, ///< [in] Pointer to the input sequence of data items
OutputIteratorT d_out, ///< [out] Pointer to the output aggregate
int num_segments, ///< [in] The number of segments that comprise the sorting data
OffsetIteratorT d_begin_offsets, ///< [in] Pointer to the sequence of beginning offsets of length \p num_segments, such that <tt>d_begin_offsets[i]</tt> is the first element of the <em>i</em><sup>th</sup> data segment in <tt>d_keys_*</tt> and <tt>d_values_*</tt>
OffsetIteratorT d_end_offsets, ///< [in] Pointer to the sequence of ending offsets of length \p num_segments, such that <tt>d_end_offsets[i]-1</tt> is the last element of the <em>i</em><sup>th</sup> data segment in <tt>d_keys_*</tt> and <tt>d_values_*</tt>. If <tt>d_end_offsets[i]-1</tt> <= <tt>d_begin_offsets[i]</tt>, the <em>i</em><sup>th</sup> is considered empty.
cudaStream_t stream = 0, ///< [in] <b>[optional]</b> CUDA stream to launch kernels within. Default is stream<sub>0</sub>.
bool debug_synchronous = false) ///< [in] <b>[optional]</b> Whether or not to synchronize the stream after every kernel launch to check for errors. Also causes launch configurations to be printed to the console. Default is \p false.
{
// Signed integer type for global offsets
typedef int OffsetT;
// The output value type
typedef typename If<(Equals<typename std::iterator_traits<OutputIteratorT>::value_type, void>::VALUE), // OutputT = (if output iterator's value type is void) ?
typename std::iterator_traits<InputIteratorT>::value_type, // ... then the input iterator's value type,
typename std::iterator_traits<OutputIteratorT>::value_type>::Type OutputT; // ... else the output iterator's value type
return DispatchSegmentedReduce<InputIteratorT, OutputIteratorT, OffsetIteratorT, OffsetT, cub::Sum>::Dispatch(
d_temp_storage,
temp_storage_bytes,
d_in,
d_out,
num_segments,
d_begin_offsets,
d_end_offsets,
cub::Sum(),
OutputT(), // zero-initialize
stream,
debug_synchronous);
}
/**
* \brief Computes a device-wide segmented minimum using the less-than ('<') operator.
*
* \par
* - Uses <tt>std::numeric_limits<T>::max()</tt> as the initial value of the reduction for each segment.
* - When input a contiguous sequence of segments, a single sequence
* \p segment_offsets (of length <tt>num_segments+1</tt>) can be aliased
* for both the \p d_begin_offsets and \p d_end_offsets parameters (where
* the latter is specified as <tt>segment_offsets+1</tt>).
* - Does not support \p < operators that are non-commutative.
* - \devicestorage
*
* \par Snippet
* The code snippet below illustrates the min-reduction of a device vector of \p int data elements.
* \par
* \code
* #include <cub/cub.cuh> // or equivalently <cub/device/device_radix_sort.cuh>
*
* // Declare, allocate, and initialize device-accessible pointers for input and output
* int num_segments; // e.g., 3
* int *d_offsets; // e.g., [0, 3, 3, 7]
* int *d_in; // e.g., [8, 6, 7, 5, 3, 0, 9]
* int *d_out; // e.g., [-, -, -]
* ...
*
* // Determine temporary device storage requirements
* void *d_temp_storage = NULL;
* size_t temp_storage_bytes = 0;
* cub::DeviceSegmentedReduce::Min(d_temp_storage, temp_storage_bytes, d_in, d_out,
* num_segments, d_offsets, d_offsets + 1);
*
* // Allocate temporary storage
* cudaMalloc(&d_temp_storage, temp_storage_bytes);
*
* // Run min-reduction
* cub::DeviceSegmentedReduce::Min(d_temp_storage, temp_storage_bytes, d_in, d_out,
* num_segments, d_offsets, d_offsets + 1);
*
* // d_out <-- [6, INT_MAX, 0]
*
* \endcode
*
* \tparam InputIteratorT <b>[inferred]</b> Random-access input iterator type for reading input items \iterator
* \tparam OutputIteratorT <b>[inferred]</b> Output iterator type for recording the reduced aggregate \iterator
* \tparam OffsetIteratorT <b>[inferred]</b> Random-access input iterator type for reading segment offsets \iterator
*/
template <
typename InputIteratorT,
typename OutputIteratorT,
typename OffsetIteratorT>
CUB_RUNTIME_FUNCTION
static cudaError_t Min(
void *d_temp_storage, ///< [in] %Device-accessible allocation of temporary storage. When NULL, the required allocation size is written to \p temp_storage_bytes and no work is done.
size_t &temp_storage_bytes, ///< [in,out] Reference to size in bytes of \p d_temp_storage allocation
InputIteratorT d_in, ///< [in] Pointer to the input sequence of data items
OutputIteratorT d_out, ///< [out] Pointer to the output aggregate
int num_segments, ///< [in] The number of segments that comprise the sorting data
OffsetIteratorT d_begin_offsets, ///< [in] Pointer to the sequence of beginning offsets of length \p num_segments, such that <tt>d_begin_offsets[i]</tt> is the first element of the <em>i</em><sup>th</sup> data segment in <tt>d_keys_*</tt> and <tt>d_values_*</tt>
OffsetIteratorT d_end_offsets, ///< [in] Pointer to the sequence of ending offsets of length \p num_segments, such that <tt>d_end_offsets[i]-1</tt> is the last element of the <em>i</em><sup>th</sup> data segment in <tt>d_keys_*</tt> and <tt>d_values_*</tt>. If <tt>d_end_offsets[i]-1</tt> <= <tt>d_begin_offsets[i]</tt>, the <em>i</em><sup>th</sup> is considered empty.
cudaStream_t stream = 0, ///< [in] <b>[optional]</b> CUDA stream to launch kernels within. Default is stream<sub>0</sub>.
bool debug_synchronous = false) ///< [in] <b>[optional]</b> Whether or not to synchronize the stream after every kernel launch to check for errors. Also causes launch configurations to be printed to the console. Default is \p false.
{
// Signed integer type for global offsets
typedef int OffsetT;
// The input value type
typedef typename std::iterator_traits<InputIteratorT>::value_type InputT;
return DispatchSegmentedReduce<InputIteratorT, OutputIteratorT, OffsetIteratorT, OffsetT, cub::Min>::Dispatch(
d_temp_storage,
temp_storage_bytes,
d_in,
d_out,
num_segments,
d_begin_offsets,
d_end_offsets,
cub::Min(),
Traits<InputT>::Max(), // replace with std::numeric_limits<T>::max() when C++11 support is more prevalent
stream,
debug_synchronous);
}
/**
* \brief Finds the first device-wide minimum in each segment using the less-than ('<') operator, also returning the in-segment index of that item.
*
* \par
* - The output value type of \p d_out is cub::KeyValuePair <tt><int, T></tt> (assuming the value type of \p d_in is \p T)
* - The minimum of the <em>i</em><sup>th</sup> segment is written to <tt>d_out[i].value</tt> and its offset in that segment is written to <tt>d_out[i].key</tt>.
* - The <tt>{1, std::numeric_limits<T>::max()}</tt> tuple is produced for zero-length inputs
* - When input a contiguous sequence of segments, a single sequence
* \p segment_offsets (of length <tt>num_segments+1</tt>) can be aliased
* for both the \p d_begin_offsets and \p d_end_offsets parameters (where
* the latter is specified as <tt>segment_offsets+1</tt>).
* - Does not support \p < operators that are non-commutative.
* - \devicestorage
*
* \par Snippet
* The code snippet below illustrates the argmin-reduction of a device vector of \p int data elements.
* \par
* \code
* #include <cub/cub.cuh> // or equivalently <cub/device/device_radix_sort.cuh>
*
* // Declare, allocate, and initialize device-accessible pointers for input and output
* int num_segments; // e.g., 3
* int *d_offsets; // e.g., [0, 3, 3, 7]
* int *d_in; // e.g., [8, 6, 7, 5, 3, 0, 9]
* KeyValuePair<int, int> *d_out; // e.g., [{-,-}, {-,-}, {-,-}]
* ...
*
* // Determine temporary device storage requirements
* void *d_temp_storage = NULL;
* size_t temp_storage_bytes = 0;
* cub::DeviceSegmentedReduce::ArgMin(d_temp_storage, temp_storage_bytes, d_in, d_out,
* num_segments, d_offsets, d_offsets + 1);
*
* // Allocate temporary storage
* cudaMalloc(&d_temp_storage, temp_storage_bytes);
*
* // Run argmin-reduction
* cub::DeviceSegmentedReduce::ArgMin(d_temp_storage, temp_storage_bytes, d_in, d_out,
* num_segments, d_offsets, d_offsets + 1);
*
* // d_out <-- [{1,6}, {1,INT_MAX}, {2,0}]
*
* \endcode
*
* \tparam InputIteratorT <b>[inferred]</b> Random-access input iterator type for reading input items (of some type \p T) \iterator
* \tparam OutputIteratorT <b>[inferred]</b> Output iterator type for recording the reduced aggregate (having value type <tt>KeyValuePair<int, T></tt>) \iterator
* \tparam OffsetIteratorT <b>[inferred]</b> Random-access input iterator type for reading segment offsets \iterator
*/
template <
typename InputIteratorT,
typename OutputIteratorT,
typename OffsetIteratorT>
CUB_RUNTIME_FUNCTION
static cudaError_t ArgMin(
void *d_temp_storage, ///< [in] %Device-accessible allocation of temporary storage. When NULL, the required allocation size is written to \p temp_storage_bytes and no work is done.
size_t &temp_storage_bytes, ///< [in,out] Reference to size in bytes of \p d_temp_storage allocation
InputIteratorT d_in, ///< [in] Pointer to the input sequence of data items
OutputIteratorT d_out, ///< [out] Pointer to the output aggregate
int num_segments, ///< [in] The number of segments that comprise the sorting data
OffsetIteratorT d_begin_offsets, ///< [in] Pointer to the sequence of beginning offsets of length \p num_segments, such that <tt>d_begin_offsets[i]</tt> is the first element of the <em>i</em><sup>th</sup> data segment in <tt>d_keys_*</tt> and <tt>d_values_*</tt>
OffsetIteratorT d_end_offsets, ///< [in] Pointer to the sequence of ending offsets of length \p num_segments, such that <tt>d_end_offsets[i]-1</tt> is the last element of the <em>i</em><sup>th</sup> data segment in <tt>d_keys_*</tt> and <tt>d_values_*</tt>. If <tt>d_end_offsets[i]-1</tt> <= <tt>d_begin_offsets[i]</tt>, the <em>i</em><sup>th</sup> is considered empty.
cudaStream_t stream = 0, ///< [in] <b>[optional]</b> CUDA stream to launch kernels within. Default is stream<sub>0</sub>.
bool debug_synchronous = false) ///< [in] <b>[optional]</b> Whether or not to synchronize the stream after every kernel launch to check for errors. Also causes launch configurations to be printed to the console. Default is \p false.
{
// Signed integer type for global offsets
typedef int OffsetT;
// The input type
typedef typename std::iterator_traits<InputIteratorT>::value_type InputValueT;
// The output tuple type
typedef typename If<(Equals<typename std::iterator_traits<OutputIteratorT>::value_type, void>::VALUE), // OutputT = (if output iterator's value type is void) ?
KeyValuePair<OffsetT, InputValueT>, // ... then the key value pair OffsetT + InputValueT
typename std::iterator_traits<OutputIteratorT>::value_type>::Type OutputTupleT; // ... else the output iterator's value type
// The output value type
typedef typename OutputTupleT::Value OutputValueT;
// Wrapped input iterator to produce index-value <OffsetT, InputT> tuples
typedef ArgIndexInputIterator<InputIteratorT, OffsetT, OutputValueT> ArgIndexInputIteratorT;
ArgIndexInputIteratorT d_indexed_in(d_in);
// Initial value
OutputTupleT initial_value(1, Traits<InputValueT>::Max()); // replace with std::numeric_limits<T>::max() when C++11 support is more prevalent
return DispatchSegmentedReduce<ArgIndexInputIteratorT, OutputIteratorT, OffsetIteratorT, OffsetT, cub::ArgMin>::Dispatch(
d_temp_storage,
temp_storage_bytes,
d_indexed_in,
d_out,
num_segments,
d_begin_offsets,
d_end_offsets,
cub::ArgMin(),
initial_value,
stream,
debug_synchronous);
}
/**
* \brief Computes a device-wide segmented maximum using the greater-than ('>') operator.
*
* \par
* - Uses <tt>std::numeric_limits<T>::lowest()</tt> as the initial value of the reduction.
* - When input a contiguous sequence of segments, a single sequence
* \p segment_offsets (of length <tt>num_segments+1</tt>) can be aliased
* for both the \p d_begin_offsets and \p d_end_offsets parameters (where
* the latter is specified as <tt>segment_offsets+1</tt>).
* - Does not support \p > operators that are non-commutative.
* - \devicestorage
*
* \par Snippet
* The code snippet below illustrates the max-reduction of a device vector of \p int data elements.
* \par
* \code
* #include <cub/cub.cuh> // or equivalently <cub/device/device_radix_sort.cuh>
*
* // Declare, allocate, and initialize device-accessible pointers for input and output
* int num_segments; // e.g., 3
* int *d_offsets; // e.g., [0, 3, 3, 7]
* int *d_in; // e.g., [8, 6, 7, 5, 3, 0, 9]
* int *d_out; // e.g., [-, -, -]
* ...
*
* // Determine temporary device storage requirements
* void *d_temp_storage = NULL;
* size_t temp_storage_bytes = 0;
* cub::DeviceSegmentedReduce::Max(d_temp_storage, temp_storage_bytes, d_in, d_out,
* num_segments, d_offsets, d_offsets + 1);
*
* // Allocate temporary storage
* cudaMalloc(&d_temp_storage, temp_storage_bytes);
*
* // Run max-reduction
* cub::DeviceSegmentedReduce::Max(d_temp_storage, temp_storage_bytes, d_in, d_out,
* num_segments, d_offsets, d_offsets + 1);
*
* // d_out <-- [8, INT_MIN, 9]
*
* \endcode
*
* \tparam InputIteratorT <b>[inferred]</b> Random-access input iterator type for reading input items \iterator
* \tparam OutputIteratorT <b>[inferred]</b> Output iterator type for recording the reduced aggregate \iterator
* \tparam OffsetIteratorT <b>[inferred]</b> Random-access input iterator type for reading segment offsets \iterator
*/
template <
typename InputIteratorT,
typename OutputIteratorT,
typename OffsetIteratorT>
CUB_RUNTIME_FUNCTION
static cudaError_t Max(
void *d_temp_storage, ///< [in] %Device-accessible allocation of temporary storage. When NULL, the required allocation size is written to \p temp_storage_bytes and no work is done.
size_t &temp_storage_bytes, ///< [in,out] Reference to size in bytes of \p d_temp_storage allocation
InputIteratorT d_in, ///< [in] Pointer to the input sequence of data items
OutputIteratorT d_out, ///< [out] Pointer to the output aggregate
int num_segments, ///< [in] The number of segments that comprise the sorting data
OffsetIteratorT d_begin_offsets, ///< [in] Pointer to the sequence of beginning offsets of length \p num_segments, such that <tt>d_begin_offsets[i]</tt> is the first element of the <em>i</em><sup>th</sup> data segment in <tt>d_keys_*</tt> and <tt>d_values_*</tt>
OffsetIteratorT d_end_offsets, ///< [in] Pointer to the sequence of ending offsets of length \p num_segments, such that <tt>d_end_offsets[i]-1</tt> is the last element of the <em>i</em><sup>th</sup> data segment in <tt>d_keys_*</tt> and <tt>d_values_*</tt>. If <tt>d_end_offsets[i]-1</tt> <= <tt>d_begin_offsets[i]</tt>, the <em>i</em><sup>th</sup> is considered empty.
cudaStream_t stream = 0, ///< [in] <b>[optional]</b> CUDA stream to launch kernels within. Default is stream<sub>0</sub>.
bool debug_synchronous = false) ///< [in] <b>[optional]</b> Whether or not to synchronize the stream after every kernel launch to check for errors. Also causes launch configurations to be printed to the console. Default is \p false.
{
// Signed integer type for global offsets
typedef int OffsetT;
// The input value type
typedef typename std::iterator_traits<InputIteratorT>::value_type InputT;
return DispatchSegmentedReduce<InputIteratorT, OutputIteratorT, OffsetIteratorT, OffsetT, cub::Max>::Dispatch(
d_temp_storage,
temp_storage_bytes,
d_in,
d_out,
num_segments,
d_begin_offsets,
d_end_offsets,
cub::Max(),
Traits<InputT>::Lowest(), // replace with std::numeric_limits<T>::lowest() when C++11 support is more prevalent
stream,
debug_synchronous);
}
/**
* \brief Finds the first device-wide maximum in each segment using the greater-than ('>') operator, also returning the in-segment index of that item
*
* \par
* - The output value type of \p d_out is cub::KeyValuePair <tt><int, T></tt> (assuming the value type of \p d_in is \p T)
* - The maximum of the <em>i</em><sup>th</sup> segment is written to <tt>d_out[i].value</tt> and its offset in that segment is written to <tt>d_out[i].key</tt>.
* - The <tt>{1, std::numeric_limits<T>::lowest()}</tt> tuple is produced for zero-length inputs
* - When input a contiguous sequence of segments, a single sequence
* \p segment_offsets (of length <tt>num_segments+1</tt>) can be aliased
* for both the \p d_begin_offsets and \p d_end_offsets parameters (where
* the latter is specified as <tt>segment_offsets+1</tt>).
* - Does not support \p > operators that are non-commutative.
* - \devicestorage
*
* \par Snippet
* The code snippet below illustrates the argmax-reduction of a device vector of \p int data elements.
* \par
* \code
* #include <cub/cub.cuh> // or equivalently <cub/device/device_reduce.cuh>
*
* // Declare, allocate, and initialize device-accessible pointers for input and output
* int num_segments; // e.g., 3
* int *d_offsets; // e.g., [0, 3, 3, 7]
* int *d_in; // e.g., [8, 6, 7, 5, 3, 0, 9]
* KeyValuePair<int, int> *d_out; // e.g., [{-,-}, {-,-}, {-,-}]
* ...
*
* // Determine temporary device storage requirements
* void *d_temp_storage = NULL;
* size_t temp_storage_bytes = 0;
* cub::DeviceSegmentedReduce::ArgMax(d_temp_storage, temp_storage_bytes, d_in, d_out,
* num_segments, d_offsets, d_offsets + 1);
*
* // Allocate temporary storage
* cudaMalloc(&d_temp_storage, temp_storage_bytes);
*
* // Run argmax-reduction
* cub::DeviceSegmentedReduce::ArgMax(d_temp_storage, temp_storage_bytes, d_in, d_out,
* num_segments, d_offsets, d_offsets + 1);
*
* // d_out <-- [{0,8}, {1,INT_MIN}, {3,9}]
*
* \endcode
*
* \tparam InputIteratorT <b>[inferred]</b> Random-access input iterator type for reading input items (of some type \p T) \iterator
* \tparam OutputIteratorT <b>[inferred]</b> Output iterator type for recording the reduced aggregate (having value type <tt>KeyValuePair<int, T></tt>) \iterator
* \tparam OffsetIteratorT <b>[inferred]</b> Random-access input iterator type for reading segment offsets \iterator
*/
template <
typename InputIteratorT,
typename OutputIteratorT,
typename OffsetIteratorT>
CUB_RUNTIME_FUNCTION
static cudaError_t ArgMax(
void *d_temp_storage, ///< [in] %Device-accessible allocation of temporary storage. When NULL, the required allocation size is written to \p temp_storage_bytes and no work is done.
size_t &temp_storage_bytes, ///< [in,out] Reference to size in bytes of \p d_temp_storage allocation
InputIteratorT d_in, ///< [in] Pointer to the input sequence of data items
OutputIteratorT d_out, ///< [out] Pointer to the output aggregate
int num_segments, ///< [in] The number of segments that comprise the sorting data
OffsetIteratorT d_begin_offsets, ///< [in] Pointer to the sequence of beginning offsets of length \p num_segments, such that <tt>d_begin_offsets[i]</tt> is the first element of the <em>i</em><sup>th</sup> data segment in <tt>d_keys_*</tt> and <tt>d_values_*</tt>
OffsetIteratorT d_end_offsets, ///< [in] Pointer to the sequence of ending offsets of length \p num_segments, such that <tt>d_end_offsets[i]-1</tt> is the last element of the <em>i</em><sup>th</sup> data segment in <tt>d_keys_*</tt> and <tt>d_values_*</tt>. If <tt>d_end_offsets[i]-1</tt> <= <tt>d_begin_offsets[i]</tt>, the <em>i</em><sup>th</sup> is considered empty.
cudaStream_t stream = 0, ///< [in] <b>[optional]</b> CUDA stream to launch kernels within. Default is stream<sub>0</sub>.
bool debug_synchronous = false) ///< [in] <b>[optional]</b> Whether or not to synchronize the stream after every kernel launch to check for errors. Also causes launch configurations to be printed to the console. Default is \p false.
{
// Signed integer type for global offsets
typedef int OffsetT;
// The input type
typedef typename std::iterator_traits<InputIteratorT>::value_type InputValueT;
// The output tuple type
typedef typename If<(Equals<typename std::iterator_traits<OutputIteratorT>::value_type, void>::VALUE), // OutputT = (if output iterator's value type is void) ?
KeyValuePair<OffsetT, InputValueT>, // ... then the key value pair OffsetT + InputValueT
typename std::iterator_traits<OutputIteratorT>::value_type>::Type OutputTupleT; // ... else the output iterator's value type
// The output value type
typedef typename OutputTupleT::Value OutputValueT;
// Wrapped input iterator to produce index-value <OffsetT, InputT> tuples
typedef ArgIndexInputIterator<InputIteratorT, OffsetT, OutputValueT> ArgIndexInputIteratorT;
ArgIndexInputIteratorT d_indexed_in(d_in);
// Initial value
OutputTupleT initial_value(1, Traits<InputValueT>::Lowest()); // replace with std::numeric_limits<T>::lowest() when C++11 support is more prevalent
return DispatchSegmentedReduce<ArgIndexInputIteratorT, OutputIteratorT, OffsetIteratorT, OffsetT, cub::ArgMax>::Dispatch(
d_temp_storage,
temp_storage_bytes,
d_indexed_in,
d_out,
num_segments,
d_begin_offsets,
d_end_offsets,
cub::ArgMax(),
initial_value,
stream,
debug_synchronous);
}
};
} // CUB namespace
CUB_NS_POSTFIX // Optional outer namespace(s)
|