custom_ops.py

from collections import defaultdict

from dataclasses import dataclass
from typing import Dict, List, Optional, Sequence, Tuple

from torchgen import dest

# disable import sorting to avoid circular dependency.
from torchgen.api.types import DispatcherSignature  # isort:skip
from torchgen.context import method_with_native_function
from torchgen.model import BackendIndex, DispatchKey, NativeFunction, Variant
from torchgen.selective_build.selector import SelectiveBuilder
from torchgen.utils import concatMap, Target


# Generates RegisterKernelStub.cpp, which provides placeholder kernels for custom operators. This will be used at
# model authoring side.
@dataclass(frozen=True)
class ComputeNativeFunctionStub:
    @method_with_native_function
    def __call__(self, f: NativeFunction) -> Optional[str]:
        if Variant.function not in f.variants:
            return None

        sig = DispatcherSignature.from_schema(
            f.func, prefix=f"wrapper_CPU_{f.func.name.overload_name}_", symint=False
        )
        assert sig is not None
        if len(f.func.returns) == 0:
            ret_name = ""
        elif len(f.func.returns) == 1:
            if f.func.arguments.out:
                ret_name = f.func.arguments.out[0].name
            else:
                ret_name = next(
                    (
                        a.name
                        for a in f.func.arguments.flat_non_out
                        if a.type == f.func.returns[0].type
                    ),
                    "",
                )
            if not ret_name:
                raise Exception(f"Can't handle this return type {f.func}")
        else:
            assert len(f.func.arguments.out) == len(f.func.returns), (
                "Out variant number of returns need to match the number of out arguments."
                f" Got outs {str(f.func.arguments.out)} but returns {str(f.func.returns)}"
            )
            # returns a tuple of out arguments
            tensor_type = "at::Tensor &"
            comma = ", "
            ret_name = f"""::std::tuple<{comma.join([tensor_type] * len(f.func.returns))}>(
                {comma.join([r.name for r in f.func.arguments.out])}
            )"""
        ret_str = f"return {ret_name};" if len(f.func.returns) > 0 else ""
        return f"""
{sig.defn()} {{
    {ret_str}
}}
    """


def gen_custom_ops_registration(
    *,
    native_functions: Sequence[NativeFunction],
    selector: SelectiveBuilder,
    backend_index: BackendIndex,
    rocm: bool,
) -> Tuple[str, str]:
    """
    Generate custom ops registration code for dest.RegisterDispatchKey.

    :param native_functions: a sequence of `NativeFunction`
    :param selector: for selective build.
    :param backend_index: kernels for all the ops.
    :param rocm: bool for dest.RegisterDispatchKey.
    :return: generated C++ code to register custom operators into PyTorch
    """
    dispatch_key = DispatchKey.CPU

    static_init_dispatch_registrations = ""
    ns_grouped_native_functions: Dict[str, List[NativeFunction]] = defaultdict(list)
    for native_function in native_functions:
        ns_grouped_native_functions[native_function.namespace].append(native_function)

    for namespace, functions in ns_grouped_native_functions.items():
        if len(functions) == 0:
            continue
        dispatch_registrations_body = "\n".join(
            list(
                concatMap(
                    dest.RegisterDispatchKey(
                        backend_index,
                        Target.REGISTRATION,
                        selector,
                        rocm=rocm,
                        symint=False,
                        class_method_name=None,
                        skip_dispatcher_op_registration=False,
                    ),
                    functions,
                )
            )
        )
        static_init_dispatch_registrations += f"""
TORCH_LIBRARY_IMPL({namespace}, {dispatch_key}, m) {{
{dispatch_registrations_body}
}};"""
    anonymous_definition = "\n".join(
        list(
            concatMap(
                dest.RegisterDispatchKey(
                    backend_index,
                    Target.ANONYMOUS_DEFINITION,
                    selector,
                    rocm=rocm,
                    symint=False,
                    class_method_name=None,
                    skip_dispatcher_op_registration=False,
                ),
                native_functions,
            )
        )
    )
    return anonymous_definition, static_init_dispatch_registrations