selected_rows_functor.cu 14.3 KB
Newer Older
1
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
2 3 4 5 6 7 8 9 10 11 12 13 14

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */

T
typhoonzero 已提交
15
#include <set>
16
#include <vector>
T
typhoonzero 已提交
17

Y
Yi Wang 已提交
18 19
#include "paddle/fluid/operators/math/math_function.h"
#include "paddle/fluid/operators/math/selected_rows_functor.h"
D
dzhwinter 已提交
20
#include "paddle/fluid/platform/cuda_primitives.h"
21 22 23 24 25

namespace paddle {
namespace operators {
namespace math {
template <typename T>
Q
QI JUN 已提交
26 27
struct SelectedRowsAdd<platform::CUDADeviceContext, T> {
  void operator()(const platform::CUDADeviceContext& context,
28 29 30 31 32 33 34
                  const framework::SelectedRows& input1,
                  const framework::SelectedRows& input2,
                  framework::SelectedRows* output) {
    auto in1_height = input1.height();
    PADDLE_ENFORCE_EQ(in1_height, input2.height());
    output->set_height(in1_height);

D
dzhwinter 已提交
35
    framework::Vector<int64_t> in1_rows(input1.rows());
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
    auto& in2_rows = input2.rows();
    std::vector<int64_t> out_rows;
    out_rows.reserve(in1_rows.size() + in2_rows.size());

    // concat rows
    out_rows.insert(out_rows.end(), in1_rows.begin(), in1_rows.end());
    out_rows.insert(out_rows.end(), in2_rows.begin(), in2_rows.end());
    output->set_rows(out_rows);

    auto* out_value = output->mutable_value();
    auto& in1_value = input1.value();
    auto& in2_value = input2.value();

    auto in1_row_numel = in1_value.numel() / in1_rows.size();
    PADDLE_ENFORCE_EQ(in1_row_numel, in2_value.numel() / in2_rows.size());
    PADDLE_ENFORCE_EQ(in1_row_numel, out_value->numel() / out_rows.size());

    auto* out_data = out_value->data<T>();
    auto* in1_data = in1_value.data<T>();

    auto in1_place = input1.place();
    PADDLE_ENFORCE(platform::is_gpu_place(in1_place));
    auto in2_place = input2.place();
    PADDLE_ENFORCE(platform::is_gpu_place(in2_place));
    auto out_place = context.GetPlace();
    PADDLE_ENFORCE(platform::is_gpu_place(out_place));

    memory::Copy(
D
dzhwinter 已提交
64 65
        boost::get<platform::CUDAPlace>(out_place), out_data,
        boost::get<platform::CUDAPlace>(in1_place), in1_data,
66 67 68 69
        in1_value.numel() * sizeof(T),
        reinterpret_cast<const platform::CUDADeviceContext&>(context).stream());

    auto* in2_data = in2_value.data<T>();
D
dzhwinter 已提交
70
    memory::Copy(boost::get<platform::CUDAPlace>(out_place),
Q
QI JUN 已提交
71
                 out_data + in1_value.numel(),
D
dzhwinter 已提交
72
                 boost::get<platform::CUDAPlace>(in2_place), in2_data,
Q
QI JUN 已提交
73
                 in2_value.numel() * sizeof(T), context.stream());
74 75 76
  }
};

Q
QI JUN 已提交
77 78
template struct SelectedRowsAdd<platform::CUDADeviceContext, float>;
template struct SelectedRowsAdd<platform::CUDADeviceContext, double>;
79 80

namespace {
Q
QI JUN 已提交
81
template <typename T, int block_size>
82 83
__global__ void SelectedRowsAddTensorKernel(const T* selected_rows,
                                            const int64_t* rows, T* tensor_out,
Q
QI JUN 已提交
84
                                            int64_t row_numel) {
85 86 87 88 89 90 91 92 93 94
  const int ty = blockIdx.y;
  int tid = threadIdx.x;

  selected_rows += ty * row_numel;
  tensor_out += rows[ty] * row_numel;

  for (int index = tid; index < row_numel; index += block_size) {
    // Since index in rows of SelectedRows can be duplicate, we can not use
    // tensor_out[index] += selected_rows[index]; Instead, we have to use
    // AtomicAdd to avoid concurrent write error.
Q
qijun 已提交
95
    paddle::platform::CudaAtomicAdd(tensor_out + index, selected_rows[index]);
96 97 98 99 100
  }
}
}  // namespace

template <typename T>
Q
QI JUN 已提交
101 102
struct SelectedRowsAddTensor<platform::CUDADeviceContext, T> {
  void operator()(const platform::CUDADeviceContext& context,
103 104 105 106 107 108 109 110 111
                  const framework::SelectedRows& input1,
                  const framework::Tensor& input2, framework::Tensor* output) {
    auto in1_height = input1.height();
    auto in2_dims = input2.dims();
    auto out_dims = output->dims();
    PADDLE_ENFORCE_EQ(in1_height, in2_dims[0]);
    PADDLE_ENFORCE_EQ(in1_height, out_dims[0]);

    auto& in1_value = input1.value();
D
dzhwinter 已提交
112
    framework::Vector<int64_t> in1_rows(input1.rows());
113 114 115 116 117 118 119 120 121

    int64_t in1_row_numel = in1_value.numel() / in1_rows.size();
    PADDLE_ENFORCE_EQ(in1_row_numel, input2.numel() / in1_height);
    PADDLE_ENFORCE_EQ(in1_row_numel, output->numel() / in1_height);

    auto* in1_data = in1_value.data<T>();
    auto* in2_data = input2.data<T>();
    auto* out_data = output->data<T>();

Q
QI JUN 已提交
122
    SetConstant<platform::CUDADeviceContext, T> functor;
123
    functor(context, output, 0.0);
124

Q
QI JUN 已提交
125
    const int block_size = 256;
126
    dim3 threads(block_size, 1);
Q
qijun 已提交
127
    dim3 grid(1, in1_rows.size());
Q
QI JUN 已提交
128 129
    SelectedRowsAddTensorKernel<
        T, block_size><<<grid, threads, 0, context.stream()>>>(
Y
Yu Yang 已提交
130 131
        in1_data, in1_rows.CUDAData(context.GetPlace()), out_data,
        in1_row_numel);
132 133 134

    auto out_eigen = framework::EigenVector<T>::Flatten(*output);
    auto in2_eigen = framework::EigenVector<T>::Flatten(input2);
Q
QI JUN 已提交
135
    out_eigen.device(*context.eigen_device()) = out_eigen + in2_eigen;
136 137 138
  }
};

Q
QI JUN 已提交
139 140
template struct SelectedRowsAddTensor<platform::CUDADeviceContext, float>;
template struct SelectedRowsAddTensor<platform::CUDADeviceContext, double>;
Q
QI JUN 已提交
141 142

template <typename T>
Q
QI JUN 已提交
143 144
struct SelectedRowsAddTo<platform::CUDADeviceContext, T> {
  void operator()(const platform::CUDADeviceContext& context,
Q
QI JUN 已提交
145 146 147 148 149 150
                  const framework::SelectedRows& input1,
                  const int64_t input2_offset,
                  framework::SelectedRows* input2) {
    auto in1_height = input1.height();
    PADDLE_ENFORCE_EQ(in1_height, input2->height());

D
dzhwinter 已提交
151
    framework::Vector<int64_t> in1_rows(input1.rows());
Q
QI JUN 已提交
152 153 154 155 156 157
    auto& in2_rows = *(input2->mutable_rows());

    auto& in1_value = input1.value();
    auto* in2_value = input2->mutable_value();

    // concat rows
Y
Fix CI  
Yu Yang 已提交
158 159 160
    if (in1_rows.size()) {
      in2_rows.Extend(in1_rows.begin(), in1_rows.end());
    }
Q
QI JUN 已提交
161 162 163 164 165 166 167 168

    auto in1_place = input1.place();
    PADDLE_ENFORCE(platform::is_gpu_place(in1_place));
    auto in2_place = input2->place();
    PADDLE_ENFORCE(platform::is_gpu_place(in2_place));

    auto* in1_data = in1_value.data<T>();
    auto* in2_data = in2_value->data<T>();
D
dzhwinter 已提交
169
    memory::Copy(boost::get<platform::CUDAPlace>(in2_place),
Q
QI JUN 已提交
170
                 in2_data + input2_offset,
D
dzhwinter 已提交
171
                 boost::get<platform::CUDAPlace>(in1_place), in1_data,
Q
QI JUN 已提交
172
                 in1_value.numel() * sizeof(T), context.stream());
Q
QI JUN 已提交
173 174 175
  }
};

Q
QI JUN 已提交
176 177 178 179
template struct SelectedRowsAddTo<platform::CUDADeviceContext, float>;
template struct SelectedRowsAddTo<platform::CUDADeviceContext, double>;
template struct SelectedRowsAddTo<platform::CUDADeviceContext, int>;
template struct SelectedRowsAddTo<platform::CUDADeviceContext, int64_t>;
Q
QI JUN 已提交
180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201

namespace {
template <typename T, int block_size>
__global__ void SelectedRowsAddToTensorKernel(const T* selected_rows,
                                              const int64_t* rows,
                                              T* tensor_out,
                                              int64_t row_numel) {
  const int ty = blockIdx.y;
  int tid = threadIdx.x;

  selected_rows += ty * row_numel;
  tensor_out += rows[ty] * row_numel;

  for (int index = tid; index < row_numel; index += block_size) {
    // Since index in rows of SelectedRows can be duplicate, we have to use
    // Atomic Operation to avoid concurrent write error.
    paddle::platform::CudaAtomicAdd(tensor_out + index, selected_rows[index]);
  }
}
}  // namespace

template <typename T>
Q
QI JUN 已提交
202 203
struct SelectedRowsAddToTensor<platform::CUDADeviceContext, T> {
  void operator()(const platform::CUDADeviceContext& context,
Q
QI JUN 已提交
204 205 206 207 208 209 210
                  const framework::SelectedRows& input1,
                  framework::Tensor* input2) {
    auto in1_height = input1.height();
    auto in2_dims = input2->dims();
    PADDLE_ENFORCE_EQ(in1_height, in2_dims[0]);

    auto& in1_value = input1.value();
D
dzhwinter 已提交
211
    framework::Vector<int64_t> in1_rows(input1.rows());
Q
QI JUN 已提交
212 213 214 215 216 217 218 219 220

    int64_t in1_row_numel = in1_value.numel() / in1_rows.size();
    PADDLE_ENFORCE_EQ(in1_row_numel, input2->numel() / in1_height);

    auto* in1_data = in1_value.data<T>();
    auto* in2_data = input2->data<T>();
    const int block_size = 256;
    dim3 threads(block_size, 1);
    dim3 grid(1, in1_rows.size());
Q
QI JUN 已提交
221 222
    SelectedRowsAddToTensorKernel<
        T, block_size><<<grid, threads, 0, context.stream()>>>(
Y
Yu Yang 已提交
223 224
        in1_data, in1_rows.CUDAData(context.GetPlace()), in2_data,
        in1_row_numel);
Q
QI JUN 已提交
225 226 227
  }
};

Q
QI JUN 已提交
228 229 230 231
template struct SelectedRowsAddToTensor<platform::CUDADeviceContext, float>;
template struct SelectedRowsAddToTensor<platform::CUDADeviceContext, double>;
template struct SelectedRowsAddToTensor<platform::CUDADeviceContext, int>;
template struct SelectedRowsAddToTensor<platform::CUDADeviceContext, int64_t>;
T
typhoonzero 已提交
232 233 234 235 236 237 238

namespace scatter {

template <typename T, int block_size>
__global__ void MergeAddKernel(const T* input, const int64_t* input_rows,
                               T* out, const int64_t* out_rows,
                               size_t out_rows_size, int64_t row_numel) {
S
sneaxiy 已提交
239
  const int ty = blockIdx.x;
T
typhoonzero 已提交
240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260
  int tid = threadIdx.x;
  __shared__ size_t out_idx;

  if (tid == 0) {
    for (size_t i = 0; i < out_rows_size; i++) {
      if (input_rows[ty] == out_rows[i]) {
        out_idx = i;
      }
    }
  }

  __syncthreads();

  input += ty * row_numel;
  out += out_idx * row_numel;
  for (int index = tid; index < row_numel; index += block_size) {
    paddle::platform::CudaAtomicAdd(out + index, input[index]);
  }
}

template <typename T>
T
typhoonzero 已提交
261 262
struct MergeAdd<platform::CUDADeviceContext, T> {
  framework::SelectedRows operator()(const platform::CUDADeviceContext& context,
T
wip  
typhoonzero 已提交
263 264
                                     const framework::SelectedRows& input) {
    framework::SelectedRows out;
S
sneaxiy 已提交
265 266 267 268 269 270 271 272
    (*this)(context, input, &out);
    return out;
  }

  void operator()(const platform::CUDADeviceContext& context,
                  const framework::SelectedRows& input,
                  framework::SelectedRows* output) {
    framework::SelectedRows& out = *output;
D
dzhwinter 已提交
273
    framework::Vector<int64_t> input_rows(input.rows());
T
typhoonzero 已提交
274 275 276 277
    std::set<int64_t> row_set(input_rows.begin(), input_rows.end());
    std::vector<int64_t> merge_rows(row_set.begin(), row_set.end());

    auto input_width = input.value().dims()[1];
T
wip  
typhoonzero 已提交
278 279 280 281

    out.set_rows(merge_rows);
    out.set_height(input.height());
    out.mutable_value()->mutable_data<T>(
T
typhoonzero 已提交
282 283 284 285 286
        framework::make_ddim(
            {static_cast<int64_t>(merge_rows.size()), input_width}),
        context.GetPlace());

    math::SetConstant<platform::CUDADeviceContext, T> constant_functor;
287
    constant_functor(context, out.mutable_value(), 0.0);
T
typhoonzero 已提交
288

T
wip  
typhoonzero 已提交
289
    auto* out_data = out.mutable_value()->data<T>();
T
typhoonzero 已提交
290 291 292 293
    auto* input_data = input.value().data<T>();

    const int block_size = 256;
    dim3 threads(block_size, 1);
S
sneaxiy 已提交
294
    dim3 grid1(input_rows.size(), 1);
T
typhoonzero 已提交
295

S
sneaxiy 已提交
296
    MergeAddKernel<T, 256><<<grid1, threads, 0, context.stream()>>>(
Y
Yu Yang 已提交
297 298 299
        input_data, input_rows.CUDAData(context.GetPlace()), out_data,
        out.mutable_rows()->CUDAMutableData(context.GetPlace()),
        out.rows().size(), input_width);
T
typhoonzero 已提交
300 301 302
  }
};

T
typhoonzero 已提交
303 304 305 306
template struct MergeAdd<platform::CUDADeviceContext, float>;
template struct MergeAdd<platform::CUDADeviceContext, double>;
template struct MergeAdd<platform::CUDADeviceContext, int>;
template struct MergeAdd<platform::CUDADeviceContext, int64_t>;
T
wip  
typhoonzero 已提交
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

template <typename T, int block_size>
__global__ void UpdateToTensorKernel(const T* selected_rows,
                                     const int64_t* rows, const ScatterOps& op,
                                     T* tensor_out, int64_t row_numel) {
  const int ty = blockIdx.y;
  int tid = threadIdx.x;

  selected_rows += ty * row_numel;
  tensor_out += rows[ty] * row_numel;
  // FIXME(typhoonzero): use macro fix the below messy code.
  switch (op) {
    case ScatterOps::ASSIGN:
      for (int index = tid; index < row_numel; index += block_size) {
        tensor_out[index] = selected_rows[index];
      }
      break;
    case ScatterOps::ADD:
      for (int index = tid; index < row_numel; index += block_size) {
        tensor_out[index] += selected_rows[index];
      }
      break;
    case ScatterOps::SUB:
      for (int index = tid; index < row_numel; index += block_size) {
        tensor_out[index] -= selected_rows[index];
      }
      break;
    case ScatterOps::SUBBY:
      for (int index = tid; index < row_numel; index += block_size) {
        tensor_out[index] = selected_rows[index] - tensor_out[index];
      }
      break;
    case ScatterOps::MUL:
      for (int index = tid; index < row_numel; index += block_size) {
        tensor_out[index] *= selected_rows[index];
      }
      break;
    case ScatterOps::DIV:
      for (int index = tid; index < row_numel; index += block_size) {
        tensor_out[index] /= selected_rows[index];
      }
      break;
    case ScatterOps::DIVBY:
      for (int index = tid; index < row_numel; index += block_size) {
        tensor_out[index] = selected_rows[index] / tensor_out[index];
      }
      break;
  }
}

template <typename T>
T
typhoonzero 已提交
358 359 360 361
struct UpdateToTensor<platform::CUDADeviceContext, T> {
  void operator()(const platform::CUDADeviceContext& context,
                  const ScatterOps& op, const framework::SelectedRows& input1,
                  framework::Tensor* input2) {
T
wip  
typhoonzero 已提交
362 363
    // NOTE: Use SelectedRowsAddToTensor for better performance
    //       no additional MergeAdd called.
T
typhoonzero 已提交
364 365
    MergeAdd<platform::CUDADeviceContext, T> merge_func;
    auto merged_in1 = merge_func(context, input1);
T
wip  
typhoonzero 已提交
366 367 368 369 370 371 372 373 374 375 376

    auto in1_height = merged_in1.height();
    auto in2_dims = input2->dims();
    PADDLE_ENFORCE_EQ(in1_height, in2_dims[0]);

    auto& in1_value = merged_in1.value();
    auto& in1_rows = merged_in1.rows();

    int64_t in1_row_numel = in1_value.numel() / in1_rows.size();
    PADDLE_ENFORCE_EQ(in1_row_numel, input2->numel() / in1_height);

T
typhoonzero 已提交
377 378
    auto* in1_data = in1_value.template data<T>();
    auto* in2_data = input2->data<T>();
T
wip  
typhoonzero 已提交
379

T
typhoonzero 已提交
380
    dim3 threads(platform::PADDLE_CUDA_NUM_THREADS, 1);
T
wip  
typhoonzero 已提交
381
    dim3 grid(1, in1_rows.size());
T
typhoonzero 已提交
382
    UpdateToTensorKernel<T, platform::PADDLE_CUDA_NUM_THREADS><<<
D
dzhwinter 已提交
383 384
        grid, threads, 0, context.stream()>>>(in1_data, in1_rows.cuda_data(),
                                              op, in2_data, in1_row_numel);
T
wip  
typhoonzero 已提交
385 386
  }
};
T
typhoonzero 已提交
387
}  // namespace scatter
388 389 390
}  // namespace math
}  // namespace operators
}  // namespace paddle