cl_image.cpp 3.1 KB
Newer Older
Y
yangfei 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.

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 "cl_image.h"
namespace paddle_mobile {
L
liuruilong 已提交
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
namespace framework {
void CLImageToTensor(CLImage *cl_image, Tensor *tensor,
                     cl_command_queue commandQueue) {
  DDim ddim = cl_image->dims();
  size_t N, C, H, W;
  if (ddim.size() == 4) {
    N = ddim[0];
    if (N < 0) {
      N = 1;
    }
    C = ddim[1];
    H = ddim[2];
    W = ddim[3];
  } else if (ddim.size() == 1) {
    N = 1;
    C = ddim[0];
    H = 1;
    W = 1;
  }

  size_t width = W * ((C + 3) / 4);
  size_t height = H * N;

  float *p = tensor->data<float>();
  half imageData[width * height * 4];
  cl_int err;
  cl_mem image = cl_image->GetCLImage();
  size_t origin[3] = {0, 0, 0};
  size_t region[3] = {width, height, 1};
  err = clEnqueueReadImage(commandQueue, image, CL_TRUE, origin, region, 0, 0,
                           imageData, 0, NULL, NULL);
  size_t i0 = 0;
  for (int n = 0; n < N; n++) {
    for (int c = 0; c < C; c++) {
      size_t i1 = i0;
      for (int h = 0; h < H; h++) {
        size_t i2 = (i1 << 2) + c % 4;
        for (int w = 0; w < W; w++) {
L
liuruilong 已提交
55
          *p = Half2Float(imageData[i2]);
L
liuruilong 已提交
56 57
          i2 += 4;
          p++;
Y
yangfei 已提交
58
        }
L
liuruilong 已提交
59 60 61 62 63
        i1 += width;
      }
    }
    i0 += width * H;
  }
Y
yangfei 已提交
64

L
liuruilong 已提交
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
  if (err != CL_SUCCESS) {
    // TODO: error handling
  }
}
void TensorToCLImage(const Tensor *tensor, CLImage *cl_image,
                     cl_command_queue commandQueue) {
  DDim ddim = cl_image->dims();
  size_t N, C, H, W;
  if (ddim.size() == 4) {
    N = ddim[0];
    if (N < 0) {
      N = 1;
    }
    C = ddim[1];
    H = ddim[2];
    W = ddim[3];
  } else if (ddim.size() == 1) {
    N = 1;
    C = ddim[0];
    H = 1;
    W = 1;
  }

  size_t width = W * ((C + 3) / 4);
  size_t height = H * N;

  const float *p = tensor->data<float>();
  half imageData[width * height * 4];
  cl_mem image = cl_image->GetCLImage();
  size_t origin[3] = {0, 0, 0};
  size_t region[3] = {width, height, 1};
  cl_int err;
  err = clEnqueueReadImage(commandQueue, image, CL_TRUE, origin, region, 0, 0,
                           imageData, 0, NULL, NULL);
  if (err != CL_SUCCESS) {
    // TODO: error handling
  }
  size_t i0 = 0;
  for (int n = 0; n < N; n++) {
    for (int c = 0; c < C; c++) {
      size_t i1 = i0;
      for (int h = 0; h < H; h++) {
        size_t i2 = (i1 << 2) + c % 4;
        for (int w = 0; w < W; w++) {
L
liuruilong 已提交
109
          imageData[i2] = Float2Half(*p);
L
liuruilong 已提交
110 111
          i2 += 4;
          p++;
Y
yangfei 已提交
112
        }
L
liuruilong 已提交
113 114
        i1 += width;
      }
Y
yangfei 已提交
115
    }
L
liuruilong 已提交
116 117
    i0 += width * H;
  }
Y
yangfei 已提交
118
}
L
liuruilong 已提交
119 120
}  // namespace framework
}  // namespace paddle_mobile