flopscope.accounting.reduction

flopscope.accounting.reduction_cost(op_name: 'str', *, input_shape: 'tuple[int, ...]', axis: 'int | None' = None, symmetry: 'SymmetryGroup | None' = None) -> 'int'[flopscope source]

Weighted FLOP cost of a reduction operation.

Parameters

op_name:str: Operation name used for weight lookup, e.g. "sum" or "max".
input_shape:tuple of int: Shape of the reduction input.
axis:int or None, optional: Reduction axis. Accepted for API consistency with the analytical helper.
symmetry:SymmetryGroup or None, optional: If provided, only unique elements are counted analytically before the operation weight is applied.

Returns

:int: Weighted public cost estimate, floored to match runtime accounting.

Notes

This helper multiplies the analytical reduction count by the active weight for op_name and then floors with int(...) to match runtime budget accounting.

Examples

>>> import flopscope as flops
>>> cost = flops.accounting.reduction_cost(
... "sum",
... input_shape=(4, 5),
... axis=1,
... )
>>> isinstance(cost, int)
True