Module simd 

SIMD-optimized distance computation

This module provides platform-specific SIMD implementations for common distance metrics used in vector search. Supports:

• ARM NEON (aarch64)
• x86 SSE/AVX/AVX2/AVX-512
• Scalar fallback for other platforms

Distance metrics:

• L2 (Euclidean) distance
• Cosine similarity/distance
• Dot product

The implementations use runtime feature detection to select the best available instruction set for the current CPU.

Functions

cosine_distance

Compute cosine distance (1 - cosine similarity) using SIMD

dot_product

Compute dot product between two vectors using SIMD

l2_distance

Compute L2 (Euclidean) distance between two vectors using SIMD