#include "op_plugin/AclOpsInterface.h"
#include "op_plugin/OpApiInterface.h"
#include "op_plugin/utils/op_api_common.h"
namespace op_api {
using npu_preparation = at_npu::native::OpPreparation;
* Different from Pytorch1.11 for torch.floor_divide() using truncation division,
* hostapi are corrected to use floor division.
*/
static at::Tensor& floor_divide_out_npu_opapi(const at::Tensor& self, const at::Tensor& other, at::Tensor& result)
{
if (other.dim() == 0 && !torch_npu::utils::is_npu(other)) {
c10::Scalar others = other.item();
EXEC_NPU_CMD(aclnnFloorDivides, self, others, result);
} else {
EXEC_NPU_CMD(aclnnFloorDivide, self, other, result);
}
return result;
}
static at::Tensor self_tensor_to_device(const at::Tensor& tensor, const at::ScalarType result_type,
const c10::Device device)
{
if (npu_preparation::is_scalar_wrapped_to_tensor(tensor)) {
at::Scalar scalar = tensor.item();
return npu_preparation::copy_scalar_to_device(scalar, result_type, device);
}
return tensor;
}
static at::Tensor& inplace_floor_divide_out_npu_opapi(at::Tensor& self, const at::Tensor& other)
{
if (other.dim() == 0 && !torch_npu::utils::is_npu(other)) {
c10::Scalar others = other.item();
EXEC_NPU_CMD(aclnnInplaceFloorDivides, self, others);
} else {
EXEC_NPU_CMD(aclnnInplaceFloorDivide, self, other);
}
return self;
}
at::Tensor& floor_divide_out(const at::Tensor& self, const at::Tensor& other, at::Tensor& out)
{
DO_COMPATIBILITY(aclnnFloorDivides, acl_op::floor_divide_out(self, other, out));
DO_COMPATIBILITY(aclnnFloorDivide, acl_op::floor_divide_out(self, other, out));
std::vector<at::Tensor> tensor_list = {self, other};
auto maybe_names = op_plugin::utils::compute_names_npu(tensor_list);
auto output_size = op_infer::broadcast_ops_npu_output_size(self, other);
at::ScalarType result_type = at::native::result_type(self, other);
at::Tensor self_cp = self_tensor_to_device(self, result_type, out.device());
npu_preparation::check_tensor({self, other}, out, out.scalar_type(), output_size);
floor_divide_out_npu_opapi(self_cp, other, out);
at::namedinference::propagate_names_if_nonempty(out, maybe_names);
return out;
}
at::Tensor floor_divide(const at::Tensor& self, const at::Tensor& other)
{
DO_COMPATIBILITY(aclnnFloorDivides, acl_op::floor_divide(self, other));
DO_COMPATIBILITY(aclnnFloorDivide, acl_op::floor_divide(self, other));
std::vector<at::Tensor> tensor_list = {self, other};
auto maybe_names = op_plugin::utils::compute_names_npu(tensor_list);
bool isSelfWrapped = npu_preparation::is_scalar_wrapped_to_tensor(self);
at::Tensor outputTensor = isSelfWrapped ? other : self;
auto outputSize = op_infer::broadcast_ops_npu_output_size(self, other);
at::ScalarType high_type = at::native::result_type(self, other);
at::Tensor self_cp = self_tensor_to_device(self, high_type, outputTensor.device());
at::Tensor result = npu_preparation::apply_tensor_without_format(outputSize, outputTensor.options().dtype(high_type));
floor_divide_out_npu_opapi(self_cp, other, result);
at::namedinference::propagate_names_if_nonempty(result, maybe_names);
return result;
}
at::Tensor floor_divide(const at::Tensor& self, const at::Scalar& other)
{
DO_COMPATIBILITY(aclnnFloorDivides, acl_op::floor_divide(self, other));
auto outputSize = op_infer::input_same_output_size(self);
at::ScalarType high_type = at::native::result_type(self, other);
at::Tensor result = npu_preparation::apply_tensor_without_format(outputSize, self.options().dtype(high_type));
EXEC_NPU_CMD(aclnnFloorDivides, self, other, result);
at::namedinference::propagate_names(result, self);
return result;
}
at::Tensor& floor_divide_(at::Tensor& self, const at::Tensor& other)
{
DO_COMPATIBILITY(aclnnInplaceFloorDivides, acl_op::floor_divide_(self, other));
DO_COMPATIBILITY(aclnnInplaceFloorDivide, acl_op::floor_divide_(self, other));
std::vector<at::Tensor> tensor_list = {self, other};
auto maybe_names = op_plugin::utils::compute_names_npu(tensor_list);
npu_preparation::CheckMemory({self, other}, {self});
inplace_floor_divide_out_npu_opapi(self, other);
at::namedinference::propagate_names_if_nonempty(self, maybe_names);
return self;
}
at::Tensor& floor_divide_(at::Tensor& self, const at::Scalar& other)
{
DO_COMPATIBILITY(aclnnInplaceFloorDivides, acl_op::floor_divide_(self, other));
EXEC_NPU_CMD(aclnnInplaceFloorDivides, self, other);
return self;
}
}