#include <torch/library.h>

#include "registration.h"
#include "torch_binding.h"

TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
  ops.def("fhe_ntt_forward(Tensor(a!) a, Tensor tw, Tensor tw_shoup, Tensor primes) -> ()");
  ops.def("fhe_ntt_inverse(Tensor(a!) a, Tensor itw, Tensor itw_shoup, Tensor ninv, "
          "Tensor ninv_shoup, Tensor primes) -> ()");
  ops.def("fhe_pointwise_mul(Tensor(a!) out, Tensor a, Tensor b, Tensor primes) -> ()");

#if defined(CUDA_KERNEL) || defined(ROCM_KERNEL)
  ops.impl("fhe_ntt_forward", torch::kCUDA, &fhe_ntt_forward);
  ops.impl("fhe_ntt_inverse", torch::kCUDA, &fhe_ntt_inverse);
  ops.impl("fhe_pointwise_mul", torch::kCUDA, &fhe_pointwise_mul);
#endif
}

REGISTER_EXTENSION(TORCH_EXTENSION_NAME)