Function TEST

tensorflow/core/kernels/eigen_spatial_convolutions_test.cc:29–72

static int ceil_div(int a, int b) { return (a + b - 1) / b; }

TEST(EigenSpatialConvolutionsTest, Simple) {
  const int input_depth = 7;
  const int input_rows = 4;
  const int input_cols = 5;
  const int output_depth = 10;
  const int patch_rows = 3;
  const int patch_cols = 4;
  const int output_rows = input_rows;
  const int output_cols = input_cols;

  Tensor<float, 3> input(input_depth, input_rows, input_cols);
  Tensor<float, 4> kernel(output_depth, input_depth, patch_rows, patch_cols);
  Tensor<float, 3> result(output_depth, output_rows, output_cols);

  input = input.constant(11.0f) + input.random();
  kernel = kernel.constant(2.0f) + kernel.random();
  result.setRandom();

  result = SpatialConvolution(input, kernel);

  EXPECT_EQ(result.dimension(0), output_depth);
  EXPECT_EQ(result.dimension(1), output_rows);
  EXPECT_EQ(result.dimension(2), output_cols);

  for (int od = 0; od < output_depth; ++od) {
    for (int i = 0; i < output_rows; ++i) {
      for (int j = 0; j < output_cols; ++j) {
        float expected = 0.0f;
        for (int c = 0; c < patch_cols; ++c) {
          for (int r = 0; r < patch_rows; ++r) {
            for (int id = 0; id < input_depth; ++id) {
              if (r - 1 + i >= 0 && c - 1 + j >= 0 && r - 1 + i < output_rows &&
                  c - 1 + j < output_cols) {
                expected +=
                    input(id, r - 1 + i, c - 1 + j) * kernel(od, id, r, c);
              }
            }
          }
        }
        EigenApprox(result(od, i, j), expected);
      }
    }
  }
}

TEST(EigenSpatialConvolutionsTest, SimpleRowMajor) {
  const int input_depth = 7;