hypot_op

→TensorImplPtr

int hypot_op(const int & a, const int & b)

source edit

Python wrappers

pylucid.ops.hypotfunction

Pythagorean length: $y = \sqrt{a^2 + b^2}$ .

Composite over square_op + add_op + sqrt_op. Naive formulation — does not perform IEEE-754 overflow-safe scaling (i.e. no $\max(|a|, |b|) \cdot \sqrt{1 + (\min / \max)^2}$ trick), matching the engine's square_op precision.

Math

y = \sqrt{a^2 + b^2}

Parameters

a, bTensorImplPtr

Broadcastable operands.

Returns

TensorImplPtr

Tensor on the broadcast shape; values $\ge 0$ .

Notes

Inputs near the float dtype's overflow threshold may lose precision — prefer a dedicated hypot kernel for safety-critical paths.