selected_rows_functor.cu 13.8 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.

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. */

#include "paddle/operators/math/math_function.h"
#include "paddle/operators/math/selected_rows_functor.h"
#include "paddle/platform/cuda_helper.h"

namespace paddle {
namespace operators {
namespace math {
template <typename T>
Q
QI JUN 已提交
23 24
struct SelectedRowsAdd<platform::CUDADeviceContext, T> {
  void operator()(const platform::CUDADeviceContext& context,
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
                  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);

    auto& in1_rows = input1.rows();
    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
61 62
        boost::get<platform::CUDAPlace>(out_place), out_data,
        boost::get<platform::CUDAPlace>(in1_place), in1_data,
63 64 65 66
        in1_value.numel() * sizeof(T),
        reinterpret_cast<const platform::CUDADeviceContext&>(context).stream());

    auto* in2_data = in2_value.data<T>();
D
dzhwinter 已提交
67
    memory::Copy(boost::get<platform::CUDAPlace>(out_place),
Q
QI JUN 已提交
68
                 out_data + in1_value.numel(),
D
dzhwinter 已提交
69
                 boost::get<platform::CUDAPlace>(in2_place), in2_data,
Q
QI JUN 已提交
70
                 in2_value.numel() * sizeof(T), context.stream());
71 72 73
  }
};

Q
QI JUN 已提交
74 75
template struct SelectedRowsAdd<platform::CUDADeviceContext, float>;
template struct SelectedRowsAdd<platform::CUDADeviceContext, double>;
76 77

namespace {
Q
QI JUN 已提交
78
template <typename T, int block_size>
79 80
__global__ void SelectedRowsAddTensorKernel(const T* selected_rows,
                                            const int64_t* rows, T* tensor_out,
Q
QI JUN 已提交
81
                                            int64_t row_numel) {
82 83 84 85 86 87 88 89 90 91
  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
92
    paddle::platform::CudaAtomicAdd(tensor_out + index, selected_rows[index]);
93 94 95 96 97
  }
}
}  // namespace

template <typename T>
Q
QI JUN 已提交
98 99
struct SelectedRowsAddTensor<platform::CUDADeviceContext, T> {
  void operator()(const platform::CUDADeviceContext& context,
100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118
                  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();
    auto& in1_rows = input1.rows();

    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
119
    SetConstant<platform::CUDADeviceContext, T> functor;
120 121
    functor(context, output, 0.0);

Q
QI JUN 已提交
122
    const int block_size = 256;
123
    dim3 threads(block_size, 1);
Q
qijun 已提交
124
    dim3 grid(1, in1_rows.size());
Q
QI JUN 已提交
125 126 127
    SelectedRowsAddTensorKernel<
        T, block_size><<<grid, threads, 0, context.stream()>>>(
        in1_data, in1_rows.data(), out_data, in1_row_numel);
128 129 130

    auto out_eigen = framework::EigenVector<T>::Flatten(*output);
    auto in2_eigen = framework::EigenVector<T>::Flatten(input2);
Q
QI JUN 已提交
131
    out_eigen.device(*context.eigen_device()) = out_eigen + in2_eigen;
132 133 134
  }
};

Q
QI JUN 已提交
135 136
template struct SelectedRowsAddTensor<platform::CUDADeviceContext, float>;
template struct SelectedRowsAddTensor<platform::CUDADeviceContext, double>;
Q
QI JUN 已提交
137 138

template <typename T>
Q
QI JUN 已提交
139 140
struct SelectedRowsAddTo<platform::CUDADeviceContext, T> {
  void operator()(const platform::CUDADeviceContext& context,
Q
QI JUN 已提交
141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162
                  const framework::SelectedRows& input1,
                  const int64_t input2_offset,
                  framework::SelectedRows* input2) {
    auto in1_height = input1.height();
    PADDLE_ENFORCE_EQ(in1_height, input2->height());

    auto& in1_rows = input1.rows();
    auto& in2_rows = *(input2->mutable_rows());

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

    // concat rows
    in2_rows.insert(in2_rows.end(), in1_rows.begin(), in1_rows.end());

    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
163
    memory::Copy(boost::get<platform::CUDAPlace>(in2_place),
Q
QI JUN 已提交
164
                 in2_data + input2_offset,
D
dzhwinter 已提交
165
                 boost::get<platform::CUDAPlace>(in1_place), in1_data,
Q
QI JUN 已提交
166
                 in1_value.numel() * sizeof(T), context.stream());
Q
QI JUN 已提交
167 168 169
  }
};

Q
QI JUN 已提交
170 171 172 173
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
174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195

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
196 197
struct SelectedRowsAddToTensor<platform::CUDADeviceContext, T> {
  void operator()(const platform::CUDADeviceContext& context,
Q
QI JUN 已提交
198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214
                  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();
    auto& in1_rows = input1.rows();

    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
215 216 217
    SelectedRowsAddToTensorKernel<
        T, block_size><<<grid, threads, 0, context.stream()>>>(
        in1_data, in1_rows.data(), in2_data, in1_row_numel);
Q
QI JUN 已提交
218 219 220
  }
};

Q
QI JUN 已提交
221 222 223 224
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
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

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) {
  const int ty = blockIdx.y;
  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>
struct MergeAdd<platform::GPUDeviceContext, T> {
T
wip  
typhoonzero 已提交
255 256 257
  framework::SelectedRows operator()(const platform::GPUDeviceContext& context,
                                     const framework::SelectedRows& input) {
    framework::SelectedRows out;
T
typhoonzero 已提交
258 259 260 261 262
    auto input_rows = input.rows();
    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
263 264 265 266

    out.set_rows(merge_rows);
    out.set_height(input.height());
    out.mutable_value()->mutable_data<T>(
T
typhoonzero 已提交
267 268 269 270 271
        framework::make_ddim(
            {static_cast<int64_t>(merge_rows.size()), input_width}),
        context.GetPlace());

    math::SetConstant<platform::CUDADeviceContext, T> constant_functor;
T
wip  
typhoonzero 已提交
272
    constant_functor(context, out.mutable_value(), 0.0);
T
typhoonzero 已提交
273

T
wip  
typhoonzero 已提交
274
    auto* out_data = out.mutable_value()->data<T>();
T
typhoonzero 已提交
275 276 277 278 279 280 281 282 283 284
    auto* input_data = input.value().data<T>();

    const int block_size = 256;
    dim3 threads(block_size, 1);
    dim3 grid1(1, input_rows.size());

    MergeAddKernel<
        T, 256><<<grid1, threads, 0,
                  reinterpret_cast<const platform::CUDADeviceContext&>(context)
                      .stream()>>>(input_data, input.rows().data(), out_data,
T
wip  
typhoonzero 已提交
285
                                   out.rows().data(), out.rows().size(),
T
typhoonzero 已提交
286
                                   input_width);
T
wip  
typhoonzero 已提交
287
    return out;
T
typhoonzero 已提交
288 289 290
  }
};

T
wip  
typhoonzero 已提交
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
template struct MergeAdd<platform::GPUDeviceContext, float>;
template struct MergeAdd<platform::GPUDeviceContext, double>;
template struct MergeAdd<platform::GPUDeviceContext, int>;
template struct MergeAdd<platform::GPUDeviceContext, int64_t>;

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>
struct UpdateToTensor<platform::GPUDeviceContext, T> {
  framework::Tensor operator()(const platform::GPUDeviceContext& context,
                               const ScatterOps& op,
                               const framework::SelectedRows& input1,
                               framework::Tensor* input2) {
    // NOTE: Use SelectedRowsAddToTensor for better performance
    //       no additional MergeAdd called.
    auto merged_in1 = MergeAdd()(context, input1);

    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);

    auto* in1_data = in1_value.data<T>();
    auto* input2_data = input2->data<T>();

    dim3 threads(PADDLE_CUDA_NUM_THREADS, 1);
    dim3 grid(1, in1_rows.size());
    UpdateToTensorKernel<
        T, PADDLE_CUDA_NUM_THREADS><<<grid, threads, 0, context.stream()>>>(
        in1_data, in1_rows.data(), op, in2_data, in1_row_numel);
  }
};
T
typhoonzero 已提交
375
}  // namespace scatter
376 377 378
}  // namespace math
}  // namespace operators
}  // namespace paddle