#include "op_plugin/AclOpsInterface.h"
#include "op_plugin/utils/OpAdapter.h"
#include "torch_npu/csrc/framework/utils/InternalFormatOpAdapter.h"
namespace acl_op {
using npu_preparation = at_npu::native::OpPreparation;
using npu_utils = at_npu::native::NpuUtils;
namespace {
at::Tensor &mv_out_npu_nocheck(at::Tensor &result, const at::Tensor &self, const at::Tensor &vec)
{
bool is_self_t = op_plugin::utils::is_transpose_last_two_dims(self);
at::Tensor contiguous_self = is_self_t ? self : npu_utils::format_contiguous(self);
at::Tensor vec_t = at::unsqueeze(vec, 1);
at_npu::native::OpCommand cmd;
cmd.Name("MatMul")
.InputWithoutContiguous(contiguous_self)
.Input(vec_t)
.Attr("transpose_x1", is_self_t)
.Attr("transpose_x2", false)
.Output(result)
.Run();
at_npu::native::npu_fast_reshape_(result);
return result;
}
}
at::Tensor &mv_out(const at::Tensor &self, const at::Tensor &vec, at::Tensor &result)
{
TORCH_CHECK(self.dim() >= 1, "mv(): input tensor must has at least 1 dimension, but got ", self.dim(),
" dimensions" + OPS_ERROR(ErrCode::PARAM));
npu_preparation::CheckOut({self}, result, self, {self.size(0)});
result.unsqueeze_(1);
if (!npu_utils::check_match(&result)) {
at::Tensor contiguous_result = npu_utils::format_contiguous(result);
mv_out_npu_nocheck(contiguous_result, self, vec);
npu_utils::format_fresh_view(result, contiguous_result);
} else {
mv_out_npu_nocheck(result, self, vec);
}
result.squeeze_(1);
return result;
}
at::Tensor mv(const at::Tensor &self, const at::Tensor &vec)
{
TORCH_CHECK(self.dim() >= 1, "mv(): input tensor must has at least 1 dimension, but got ", self.dim(),
" dimensions" + OPS_ERROR(ErrCode::PARAM));
at::Tensor result = npu_preparation::apply_tensor(self, {self.size(0), 1});
mv_out_npu_nocheck(result, self, vec);
result.squeeze_(1);
return result;
}
}