File size: 9,725 Bytes
0dc1b04 | 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 | /******************************************************************************
* Copyright (c) 2011-2021, 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 "../config.cuh"
#include "../util_type.cuh"
#include "../util_namespace.cuh"
#include "../block/block_load.cuh"
#include "../block/block_store.cuh"
#include "../block/block_adjacent_difference.cuh"
#include <thrust/system/cuda/detail/core/util.h>
CUB_NAMESPACE_BEGIN
template <
int _BLOCK_THREADS,
int _ITEMS_PER_THREAD = 1,
cub::BlockLoadAlgorithm _LOAD_ALGORITHM = cub::BLOCK_LOAD_DIRECT,
cub::CacheLoadModifier _LOAD_MODIFIER = cub::LOAD_LDG,
cub::BlockStoreAlgorithm _STORE_ALGORITHM = cub::BLOCK_STORE_DIRECT>
struct AgentAdjacentDifferencePolicy
{
static constexpr int BLOCK_THREADS = _BLOCK_THREADS;
static constexpr int ITEMS_PER_THREAD = _ITEMS_PER_THREAD;
static constexpr int ITEMS_PER_TILE = BLOCK_THREADS * ITEMS_PER_THREAD;
static constexpr cub::BlockLoadAlgorithm LOAD_ALGORITHM = _LOAD_ALGORITHM;
static constexpr cub::CacheLoadModifier LOAD_MODIFIER = _LOAD_MODIFIER;
static constexpr cub::BlockStoreAlgorithm STORE_ALGORITHM = _STORE_ALGORITHM;
};
template <typename Policy,
typename InputIteratorT,
typename OutputIteratorT,
typename DifferenceOpT,
typename OffsetT,
typename InputT,
typename OutputT,
bool MayAlias,
bool ReadLeft>
struct AgentDifference
{
using LoadIt = typename THRUST_NS_QUALIFIER::cuda_cub::core::LoadIterator<Policy, InputIteratorT>::type;
using BlockLoad = typename cub::BlockLoadType<Policy, LoadIt>::type;
using BlockStore = typename cub::BlockStoreType<Policy, OutputIteratorT, OutputT>::type;
using BlockAdjacentDifferenceT =
cub::BlockAdjacentDifference<InputT, Policy::BLOCK_THREADS>;
union _TempStorage
{
typename BlockLoad::TempStorage load;
typename BlockStore::TempStorage store;
typename BlockAdjacentDifferenceT::TempStorage adjacent_difference;
};
/// Alias wrapper allowing storage to be unioned
struct TempStorage : Uninitialized<_TempStorage> {};
static constexpr int BLOCK_THREADS = Policy::BLOCK_THREADS;
static constexpr int ITEMS_PER_THREAD = Policy::ITEMS_PER_THREAD;
static constexpr int ITEMS_PER_TILE = Policy::ITEMS_PER_TILE;
static constexpr int SHARED_MEMORY_SIZE = static_cast<int>(sizeof(TempStorage));
_TempStorage &temp_storage;
InputIteratorT input_it;
LoadIt load_it;
InputT *first_tile_previous;
OutputIteratorT result;
DifferenceOpT difference_op;
OffsetT num_items;
__device__ __forceinline__ AgentDifference(TempStorage &temp_storage,
InputIteratorT input_it,
InputT *first_tile_previous,
OutputIteratorT result,
DifferenceOpT difference_op,
OffsetT num_items)
: temp_storage(temp_storage.Alias())
, input_it(input_it)
, load_it(
THRUST_NS_QUALIFIER::cuda_cub::core::make_load_iterator(Policy(),
input_it))
, first_tile_previous(first_tile_previous)
, result(result)
, difference_op(difference_op)
, num_items(num_items)
{}
template <bool IS_LAST_TILE,
bool IS_FIRST_TILE>
__device__ __forceinline__ void consume_tile_impl(int num_remaining,
int tile_idx,
OffsetT tile_base)
{
InputT input[ITEMS_PER_THREAD];
OutputT output[ITEMS_PER_THREAD];
if (IS_LAST_TILE)
{
// Fill last elements with the first element
// because collectives are not suffix guarded
BlockLoad(temp_storage.load)
.Load(load_it + tile_base, input, num_remaining, *(load_it + tile_base));
}
else
{
BlockLoad(temp_storage.load).Load(load_it + tile_base, input);
}
CTA_SYNC();
if (ReadLeft)
{
if (IS_FIRST_TILE)
{
if (IS_LAST_TILE)
{
BlockAdjacentDifferenceT(temp_storage.adjacent_difference)
.SubtractLeftPartialTile(input,
output,
difference_op,
num_remaining);
}
else
{
BlockAdjacentDifferenceT(temp_storage.adjacent_difference)
.SubtractLeft(input, output, difference_op);
}
}
else
{
InputT tile_prev_input = MayAlias
? first_tile_previous[tile_idx]
: *(input_it + tile_base - 1);
if (IS_LAST_TILE)
{
BlockAdjacentDifferenceT(temp_storage.adjacent_difference)
.SubtractLeftPartialTile(input,
output,
difference_op,
num_remaining,
tile_prev_input);
}
else
{
BlockAdjacentDifferenceT(temp_storage.adjacent_difference)
.SubtractLeft(input, output, difference_op, tile_prev_input);
}
}
}
else
{
if (IS_LAST_TILE)
{
BlockAdjacentDifferenceT(temp_storage.adjacent_difference)
.SubtractRightPartialTile(input, output, difference_op, num_remaining);
}
else
{
InputT tile_next_input = MayAlias
? first_tile_previous[tile_idx]
: *(input_it + tile_base + ITEMS_PER_TILE);
BlockAdjacentDifferenceT(temp_storage.adjacent_difference)
.SubtractRight(input, output, difference_op, tile_next_input);
}
}
CTA_SYNC();
if (IS_LAST_TILE)
{
BlockStore(temp_storage.store)
.Store(result + tile_base, output, num_remaining);
}
else
{
BlockStore(temp_storage.store).Store(result + tile_base, output);
}
}
template <bool IS_LAST_TILE>
__device__ __forceinline__ void consume_tile(int num_remaining,
int tile_idx,
OffsetT tile_base)
{
if (tile_idx == 0)
{
consume_tile_impl<IS_LAST_TILE, true>(num_remaining,
tile_idx,
tile_base);
}
else
{
consume_tile_impl<IS_LAST_TILE, false>(num_remaining,
tile_idx,
tile_base);
}
}
__device__ __forceinline__ void Process(int tile_idx,
OffsetT tile_base)
{
OffsetT num_remaining = num_items - tile_base;
if (num_remaining > ITEMS_PER_TILE) // not a last tile
{
consume_tile<false>(num_remaining, tile_idx, tile_base);
}
else
{
consume_tile<true>(num_remaining, tile_idx, tile_base);
}
}
};
template <typename InputIteratorT,
typename InputT,
typename OffsetT,
bool ReadLeft>
struct AgentDifferenceInit
{
static constexpr int BLOCK_THREADS = 128;
static __device__ __forceinline__ void Process(int tile_idx,
InputIteratorT first,
InputT *result,
OffsetT num_tiles,
int items_per_tile)
{
OffsetT tile_base = static_cast<OffsetT>(tile_idx) * items_per_tile;
if (tile_base > 0 && tile_idx < num_tiles)
{
if (ReadLeft)
{
result[tile_idx] = first[tile_base - 1];
}
else
{
result[tile_idx - 1] = first[tile_base];
}
}
}
};
CUB_NAMESPACE_END
|