File size: 1,369 Bytes
674f9be
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
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
55
#include <torch/extension.h>
#include <vector>

// CUDA forward declarations
std::vector<torch::Tensor> corr_cuda_forward(
    torch::Tensor fmap1,
    torch::Tensor fmap2,
    torch::Tensor coords,
    int radius);

std::vector<torch::Tensor> corr_cuda_backward(
  torch::Tensor fmap1,
  torch::Tensor fmap2,
  torch::Tensor coords,
  torch::Tensor corr_grad,
  int radius);

// C++ interface
#define CHECK_CUDA(x) TORCH_CHECK(x.type().is_cuda(), #x " must be a CUDA tensor")
#define CHECK_CONTIGUOUS(x) TORCH_CHECK(x.is_contiguous(), #x " must be contiguous")
#define CHECK_INPUT(x) CHECK_CUDA(x); CHECK_CONTIGUOUS(x)

std::vector<torch::Tensor> corr_forward(
    torch::Tensor fmap1,
    torch::Tensor fmap2,
    torch::Tensor coords,
    int radius) {
  CHECK_INPUT(fmap1);
  CHECK_INPUT(fmap2);
  CHECK_INPUT(coords);

  return corr_cuda_forward(fmap1, fmap2, coords, radius);
}


std::vector<torch::Tensor> corr_backward(
    torch::Tensor fmap1,
    torch::Tensor fmap2,
    torch::Tensor coords,
    torch::Tensor corr_grad,
    int radius) {
  CHECK_INPUT(fmap1);
  CHECK_INPUT(fmap2);
  CHECK_INPUT(coords);
  CHECK_INPUT(corr_grad);

  return corr_cuda_backward(fmap1, fmap2, coords, corr_grad, radius);
}


PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
  m.def("forward", &corr_forward, "CORR forward");
  m.def("backward", &corr_backward, "CORR backward");
}