Expand description
Lane-wise conditional select and masked-negate.
Conditional SIMD blend/select operations on SimdView.
select performs a lane-wise conditional merge: for each index i,
the output is self[i] if mask[i] is true, else other[i].
§Architecture mapping
| Method | Instruction family |
|---|---|
blend (float mask) | AVX-512 _mm512_mask_blend_ps, AVX2 _mm256_blendv_ps, NEON vbslq_f32 |
| Scalar fallback | if mask[i] { self[i] } else { other[i] } |
§Zero-Cost Contract
Selection is monomorphized per (T, Arch, Align). The Align ZST governs
which load instruction is emitted; Arch is a ZST erased after codegen.