Module gf_simd 

SIMD-accelerated GF(2^16) buffer operations.

Dispatch hierarchy: AVX2 (256-bit) → SSSE3 (128-bit) → scalar.

The PSHUFB technique: decompose 16-bit GF multiply into four 4-bit lookups. VPSHUFB (AVX2) processes 32 bytes per instruction, PSHUFB (SSSE3) 16 bytes.

Additionally provides mul_add_multi which accumulates multiple source buffers × coefficients into dst in a single pass, reducing memory bandwidth by loading each dst cache line once instead of once per source.

Structs

GfMulTables

Precomputed PSHUFB tables for multiplying by a GF(2^16) constant.

Functions

mul_add_buffer

dst[i] ^= constant * src[i] for each u16 position.

mul_add_multi

Accumulate multiple source buffers into dst.

xor_buffers

XOR src into dst.