FHElium

Appearance

fhelium.backend.ntt._selection ​

Select supported native NTT schedules from represented execution facts.

algorithm_name ​

function View source

python
def algorithm_name(policy: NttBackendPolicy) -> str: ...

Identify the algorithm independently of its stage grouping.

table_layout ​

function View source

python
def table_layout(policy: NttBackendPolicy) -> str: ...

Identify interchangeable native table operands, without comparing data.

table_count ​

function View source

python
def table_count(policy: NttBackendPolicy) -> int: ...

Count table operands after the common RNS parameter Tensor.

select_policy ​

function View source

python
def select_policy(*, selected: str | None=None, algorithm: str | None=None, group_width: int | None=None, radix: int | None=None, tables: str | None=None, device_type: str | None=None, log_n: int | None=None) -> NttBackendPolicy | None: ...

Apply caller constraints and a stable default, without timing candidates.

An unspecified device alone does not select CPU or CUDA. A represented algorithm or table layout can determine a schedule before placement; the selected executor still requires its supported device at execution.

select_operation_policy ​

function View source

python
def select_operation_policy(operation, *, log_n): ...

Apply NTT constraints recorded on a concrete operation's input and attributes.