half-8


Vocabulary
math.vectors.simd

Class description
A 128-bit sequence of eight IEEE binary16 values. Elements are Factor floats; numeric stores round to nearest, ties to even, and canonicalize NaNs. Raw casts preserve bits. Arithmetic and conversions have portable implementations and optional FP16 kernels. Minimum and maximum select the number when only one operand is NaN. For zeros of different signs, minimum selects negative zero and maximum selects positive zero.

Definition


Methods