im2col.cu 13.6 KB
Newer Older
H
hedaoyuan 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14
/* 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. */

H
hedaoyuan 已提交
15 16
#include "paddle/operators/math/im2col.h"
#include "paddle/platform/cuda_helper.h"
H
hedaoyuan 已提交
17 18 19 20

namespace paddle {

template <class T>
H
hedaoyuan 已提交
21 22 23 24
__global__ void im2col(const T* data_im, int num_outs, int height, int width,
                       int filter_height, int filter_width, int stride_height,
                       int stride_width, int padding_height, int padding_width,
                       int output_height, int output_width, T* data_col) {
H
hedaoyuan 已提交
25
  int index = (blockIdx.x * gridDim.y + blockIdx.y) * blockDim.x + threadIdx.x;
H
hedaoyuan 已提交
26 27 28 29 30 31 32 33
  if (index < num_outs) {
    int w_out = index % output_width;
    index /= output_width;
    int h_out = index % output_height;
    int channel_in = index / output_height;
    int channel_out = channel_in * filter_height * filter_width;
    int h_in = h_out * stride_height;
    int w_in = w_out * stride_width;
H
hedaoyuan 已提交
34

H
hedaoyuan 已提交
35 36 37
    data_col += (channel_out * output_height + h_out) * output_width + w_out;
    for (int i = 0; i < filter_height; ++i) {
      for (int j = 0; j < filter_width; ++j) {
H
hedaoyuan 已提交
38 39
        int rIdx = int(h_in + i);
        int cIdx = int(w_in + j);
H
hedaoyuan 已提交
40 41 42 43
        if ((rIdx - (int)padding_height) >= (int)height ||
            (rIdx - (int)padding_height) < 0 ||
            (cIdx - (int)padding_width) >= (int)width ||
            (cIdx - (int)padding_width) < 0) {
H
hedaoyuan 已提交
44 45
          *data_col = 0;
        } else {
H
hedaoyuan 已提交
46 47
          rIdx = rIdx + channel_in * height - padding_height;
          cIdx = cIdx - padding_width;
H
hedaoyuan 已提交
48 49
          *data_col = data_im[rIdx * width + cIdx];
        }
H
hedaoyuan 已提交
50
        data_col += output_height * output_width;
H
hedaoyuan 已提交
51 52 53 54 55 56
      }
    }
  }
}

/*
H
hedaoyuan 已提交
57 58 59
 * im = [input_channels, input_height, input_width]
 * col =
 *   [input_channels, filter_height, filter_width, output_height, output_width]
H
hedaoyuan 已提交
60 61
 */
template <class T>
H
hedaoyuan 已提交
62
class Im2ColFunctor<kCFO, platform::GPUPlace, T> {
H
hedaoyuan 已提交
63
 public:
H
hedaoyuan 已提交
64 65 66 67 68
  void operator()(const framework::Tensor& im, framework::Tensor& col,
                  int stride_height, int stride_width, int padding_height,
                  int padding_width) {
    PADDLE_ENFORCE(im.dims().size() == 3);
    PADDLE_ENFORCE(col.dims().size() == 5);
H
hedaoyuan 已提交
69

H
hedaoyuan 已提交
70 71 72 73 74 75 76 77 78 79 80 81
    int input_channels = im.dims()[0];
    int input_height = im.dims()[1];
    int input_width = im.dims()[2];
    int filter_height = col.dims()[1];
    int filter_width = col.dims()[2];
    int output_height = col.dims()[3];
    int output_width = col.dims()[4];

    int num_outputs = input_channels * output_height * output_width;
    int blocks = (num_outputs + 1024 - 1) / 1024;
    int block_x = 512;
    int block_y = (blocks + 512 - 1) / 512;
H
hedaoyuan 已提交
82
    dim3 threads(1024, 1);
H
hedaoyuan 已提交
83 84 85 86 87
    dim3 grid(block_x, block_y);
    im2col<T><<<grid, threads>>>(
        im.data<T>(), num_outputs, input_height, input_width, filter_height,
        filter_width, stride_height, stride_width, padding_height,
        padding_width, output_height, output_width, col.data<T>());
H
hedaoyuan 已提交
88 89 90 91 92
  }
};

template <class T>
__global__ void col2im(size_t n, const T* data_col, size_t height, size_t width,
H
hedaoyuan 已提交
93 94 95 96 97
                       size_t channels, size_t filter_height,
                       size_t filter_width, size_t stride_height,
                       size_t stride_width, size_t padding_height,
                       size_t padding_width, size_t output_height,
                       size_t output_width, T* data_im) {
H
hedaoyuan 已提交
98 99 100 101 102 103 104
  size_t index =
      (blockIdx.x * gridDim.y + blockIdx.y) * blockDim.x + threadIdx.x;
  if (index < n) {
    T val = 0;
    int w = int(index % width);
    int h = int((index / width) % height);
    int c = int(index / (width * height));
H
hedaoyuan 已提交
105 106 107 108
    if ((w - (int)padding_width) >= 0 &&
        (w - (int)padding_width) < (width - 2 * padding_width) &&
        (h - (int)padding_height) >= 0 &&
        (h - padding_height) < (height - 2 * padding_height)) {
H
hedaoyuan 已提交
109
      // compute the start and end of the output
H
hedaoyuan 已提交
110 111 112 113 114 115 116 117 118
      int w_col_start = (w < (int)filter_width)
                            ? 0
                            : (w - int(filter_width)) / (int)stride_width + 1;
      int w_col_end =
          min((int)(w / (int)stride_width + 1), (int)(output_width));
      int h_col_start = (h < (int)filter_height)
                            ? 0
                            : (h - (int)filter_height) / (int)stride_height + 1;
      int h_col_end = min(int(h / stride_height + 1), int(output_height));
H
hedaoyuan 已提交
119 120 121
      for (int h_col = h_col_start; h_col < h_col_end; ++h_col) {
        for (int w_col = w_col_start; w_col < w_col_end; ++w_col) {
          // the col location: [c * width * height + h_out, w_out]
H
hedaoyuan 已提交
122 123 124 125 126
          int c_col = int(c * filter_height * filter_width) +
                      (h - h_col * (int)stride_height) * (int)filter_width +
                      (w - w_col * (int)stride_width);
          val +=
              data_col[(c_col * output_height + h_col) * output_width + w_col];
H
hedaoyuan 已提交
127 128
        }
      }
H
hedaoyuan 已提交
129 130 131 132 133
      h -= padding_height;
      w -= padding_width;
      data_im[c * ((width - 2 * padding_width) *
                   (height - 2 * padding_height)) +
              h * (width - 2 * padding_width) + w] += val;
H
hedaoyuan 已提交
134 135 136 137 138
    }
  }
}

/*
H
hedaoyuan 已提交
139 140 141
 * im = [input_channels, input_height, input_width]
 * col =
 *   [input_channels, filter_height, filter_width, output_height, output_width]
H
hedaoyuan 已提交
142 143
 */
template <class T>
H
hedaoyuan 已提交
144
class Col2ImFunctor<kCFO, platform::GPUPlace, T> {
H
hedaoyuan 已提交
145
 public:
H
hedaoyuan 已提交
146 147 148 149 150 151 152 153 154 155 156 157 158
  void operator()(framework::Tensor& im, const framework::Tensor& col,
                  int stride_height, int stride_width, int padding_height,
                  int padding_width) {
    PADDLE_ENFORCE(im.dims().size() == 3);
    PADDLE_ENFORCE(col.dims().size() == 5);

    int input_channels = im.dims()[0];
    int input_height = im.dims()[1];
    int input_width = im.dims()[2];
    int filter_height = col.dims()[1];
    int filter_width = col.dims()[2];
    int output_height = col.dims()[3];
    int output_width = col.dims()[4];
H
hedaoyuan 已提交
159

H
hedaoyuan 已提交
160 161
    size_t num_kernels = input_channels * (input_height + 2 * padding_height) *
                         (input_width + 2 * padding_width);
H
hedaoyuan 已提交
162

H
hedaoyuan 已提交
163 164 165
    size_t blocks = (num_kernels + 1024 - 1) / 1024;
    size_t block_x = 512;
    size_t block_y = (blocks + 512 - 1) / 512;
H
hedaoyuan 已提交
166
    dim3 threads(1024, 1);
H
hedaoyuan 已提交
167
    dim3 grid(block_x, block_y);
H
hedaoyuan 已提交
168 169 170

    // To avoid involving atomic operations, we will launch one kernel per
    // bottom dimension, and then in the kernel add up the top dimensions.
H
hedaoyuan 已提交
171 172 173 174 175
    col2im<T><<<grid, threads>>>(
        num_kernels, col.data<T>(), input_height + 2 * padding_height,
        input_width + 2 * padding_width, input_channels, filter_height,
        filter_width, stride_height, stride_width, padding_height,
        padding_width, output_height, output_width, im.data<T>());
H
hedaoyuan 已提交
176 177 178
  }
};

H
hedaoyuan 已提交
179 180 181 182
template class Im2ColFunctor<kCFO, platform::GPUPlace, float>;
template class Im2ColFunctor<kCFO, platform::GPUPlace, double>;
template class Col2ImFunctor<kCFO, platform::GPUPlace, float>;
template class Col2ImFunctor<kCFO, platform::GPUPlace, double>;
H
hedaoyuan 已提交
183 184

template <class T>
H
hedaoyuan 已提交
185 186 187 188 189 190 191 192 193 194 195 196 197 198 199
__global__ void im2colOCF(const T* im_data, T* col_data, int input_channels,
                          int input_height, int input_width, int filter_height,
                          int filter_width, int stride_height, int stride_width,
                          int padding_height, int padding_width,
                          int output_height, int output_width) {
  int swid = blockIdx.x;
  int shid = blockIdx.y;
  for (int channelid = threadIdx.z; channelid < input_channels;
       channelid += blockDim.z) {
    for (int idy = threadIdx.y; idy < filter_height; idy += blockDim.y) {
      for (int idx = threadIdx.x; idx < filter_width; idx += blockDim.x) {
        int width_offset = idx + swid * stride_width - padding_width;
        int height_offset = idy + shid * stride_height - padding_height;
        int im_offset = width_offset + height_offset * input_width +
                        channelid * input_height * input_width;
H
hedaoyuan 已提交
200

H
hedaoyuan 已提交
201 202 203 204
        int col_offset = idx + idy * filter_width +
                         channelid * filter_height * filter_width +
                         (shid * output_width + swid) *
                             (input_channels * filter_height * filter_width);
H
hedaoyuan 已提交
205

H
hedaoyuan 已提交
206 207 208
        if (height_offset >= input_height || height_offset < 0 ||
            width_offset >= input_width || width_offset < 0) {
          col_data[col_offset] = T(0);
H
hedaoyuan 已提交
209
        } else {
H
hedaoyuan 已提交
210
          col_data[col_offset] = im_data[im_offset];
H
hedaoyuan 已提交
211 212 213 214 215 216 217
        }
      }
    }
  }
}

/*
H
hedaoyuan 已提交
218 219 220
 * im = [input_channels, input_height, input_width]
 * col =
 *   [output_height, output_width, input_channels, filter_height, filter_width]
H
hedaoyuan 已提交
221 222
 */
template <class T>
H
hedaoyuan 已提交
223
class Im2ColFunctor<kOCF, platform::GPUPlace, T> {
H
hedaoyuan 已提交
224
 public:
H
hedaoyuan 已提交
225 226 227 228 229 230 231 232 233 234 235 236
  void operator()(const framework::Tensor& im, framework::Tensor& col,
                  int stride_height, int stride_width, int padding_height,
                  int padding_width) {
    PADDLE_ENFORCE(im.dims().size() == 3);
    PADDLE_ENFORCE(col.dims().size() == 5);
    int input_channels = im.dims()[0];
    int input_height = im.dims()[1];
    int input_width = im.dims()[2];
    int filter_height = col.dims()[3];
    int filter_width = col.dims()[4];
    int output_height = col.dims()[0];
    int output_width = col.dims()[1];
H
hedaoyuan 已提交
237

H
hedaoyuan 已提交
238 239 240 241 242 243 244 245 246 247 248
    int block_dim_x = 0;
    int block_dim_y = 0;
    if (filter_height <= 4 && filter_width <= 4) {
      block_dim_x = 4;
      block_dim_y = 4;
    } else if (filter_height <= 8 && filter_width <= 8) {
      block_dim_x = 8;
      block_dim_y = 8;
    } else if (filter_height <= 16 && filter_width <= 16) {
      block_dim_x = 16;
      block_dim_y = 16;
H
hedaoyuan 已提交
249
    } else {
H
hedaoyuan 已提交
250 251
      block_dim_x = 32;
      block_dim_y = 32;
H
hedaoyuan 已提交
252 253
    }

H
hedaoyuan 已提交
254 255 256 257 258 259 260 261
    int block_dim_z = 1024 / block_dim_x / block_dim_y;
    dim3 threads(block_dim_x, block_dim_y,
                 std::min(block_dim_z, input_channels));
    dim3 grid(output_width, output_height);
    im2colOCF<T><<<grid, threads>>>(
        im.data<T>(), col.data<T>(), input_channels, input_height, input_width,
        filter_height, filter_width, stride_height, stride_width,
        padding_height, padding_width, output_height, output_width);
H
hedaoyuan 已提交
262 263 264 265
  }
};

template <class T>
H
hedaoyuan 已提交
266 267 268 269 270 271 272 273 274 275 276 277 278 279 280
__global__ void col2imOCF(T* im_data, const T* col_data, int input_channels,
                          int input_height, int input_width, int filter_height,
                          int filter_width, int stride_height, int stride_width,
                          int padding_height, int padding_width,
                          int output_height, int output_width) {
  int swid = blockIdx.x;
  int shid = blockIdx.y;
  for (int channelid = threadIdx.z; channelid < input_channels;
       channelid += blockDim.z) {
    for (int idy = threadIdx.y; idy < filter_height; idy += blockDim.y) {
      for (int idx = threadIdx.x; idx < filter_width; idx += blockDim.x) {
        int width_offset = idx + swid * stride_width - padding_width;
        int height_offset = idy + shid * stride_height - padding_height;
        int im_offset = width_offset + height_offset * input_width +
                        channelid * input_height * input_width;
H
hedaoyuan 已提交
281

H
hedaoyuan 已提交
282 283 284 285
        int col_offset = idx + idy * filter_width +
                         channelid * filter_height * filter_width +
                         (shid * output_width + swid) *
                             (input_channels * filter_height * filter_width);
H
hedaoyuan 已提交
286

H
hedaoyuan 已提交
287 288 289 290
        if (height_offset >= 0 && height_offset < input_height &&
            width_offset >= 0 && width_offset < input_width) {
          paddle::platform::CudaAtomicAdd(im_data + im_offset,
                                          col_data[col_offset]);
H
hedaoyuan 已提交
291 292 293 294 295 296 297
        }
      }
    }
  }
}

/*
H
hedaoyuan 已提交
298 299 300
 * im = [input_channels, input_height, input_width]
 * col =
 *   [output_height, output_width, input_channels, filter_height, filter_width]
H
hedaoyuan 已提交
301 302
 */
template <class T>
H
hedaoyuan 已提交
303
class Col2ImFunctor<kOCF, platform::GPUPlace, T> {
H
hedaoyuan 已提交
304
 public:
H
hedaoyuan 已提交
305 306 307 308 309 310 311 312 313 314 315 316
  void operator()(framework::Tensor& im, const framework::Tensor& col,
                  int stride_height, int stride_width, int padding_height,
                  int padding_width) {
    PADDLE_ENFORCE(im.dims().size() == 3);
    PADDLE_ENFORCE(col.dims().size() == 5);
    int input_channels = im.dims()[0];
    int input_height = im.dims()[1];
    int input_width = im.dims()[2];
    int filter_height = col.dims()[3];
    int filter_width = col.dims()[4];
    int output_height = col.dims()[0];
    int output_width = col.dims()[1];
H
hedaoyuan 已提交
317

H
hedaoyuan 已提交
318 319 320 321 322 323 324 325 326 327 328
    int block_dim_x = 0;
    int block_dim_y = 0;
    if (filter_height <= 4 && filter_width <= 4) {
      block_dim_x = 4;
      block_dim_y = 4;
    } else if (filter_height <= 8 && filter_width <= 8) {
      block_dim_x = 8;
      block_dim_y = 8;
    } else if (filter_height <= 16 && filter_width <= 16) {
      block_dim_x = 16;
      block_dim_y = 16;
H
hedaoyuan 已提交
329
    } else {
H
hedaoyuan 已提交
330 331
      block_dim_x = 32;
      block_dim_y = 32;
H
hedaoyuan 已提交
332 333
    }

H
hedaoyuan 已提交
334 335 336 337 338 339 340 341
    int block_dim_z = 1024 / block_dim_x / block_dim_y;
    dim3 threads(block_dim_x, block_dim_y,
                 std::min(block_dim_z, input_channels));
    dim3 grid(output_width, output_height);
    col2imOCF<T><<<grid, threads, 0>>>(
        im.data<T>(), col.data<T>(), input_channels, input_height, input_width,
        filter_height, filter_width, stride_height, stride_width,
        padding_height, padding_width, output_height, output_width);
H
hedaoyuan 已提交
342 343 344
  }
};

H
hedaoyuan 已提交
345 346 347 348
template class Im2ColFunctor<kOCF, platform::GPUPlace, float>;
template class Im2ColFunctor<kOCF, platform::GPUPlace, double>;
template class Col2ImFunctor<kOCF, platform::GPUPlace, float>;
template class Col2ImFunctor<kOCF, platform::GPUPlace, double>;
H
hedaoyuan 已提交
349 350

}  // namespace paddle