Module embedding 

Embedding lookup kernel.

Matches embedding-lookup-v1.yaml. output[i] = W[token_ids[i]] — table lookup with bounds checking.

Each function provides one of three backends:

• fn embedding_scalar(...) – Pure Rust scalar reference (ground truth)
• unsafe fn embedding_avx2(...) – AVX2 SIMD implementation
• fn embedding_ptx() -> &'static str – PTX assembly source string

Functions

embedding_avx2^⚠

AVX2 embedding lookup – delegates to scalar (memory-bound, no compute).

embedding_ptx

PTX assembly for embedding lookup.

embedding_scalar

Embedding table lookup (scalar reference).