v0.2.5

v0.2.5 - 2026-03-13

Performance

• Loop Unrolling: Applied byte-level loop unrolling across three core computational bottlenecks to reduce loop control overhead and improve throughput:
  • DEFLATE Match Length: Unrolled the inner dflt() match-length finder loop by 4 bytes.
  • Adler-32 Checksum: Unrolled the inner checksum accumulation loop by 8 bytes instead of byte-by-byte iteration, reducing loop condition evaluations per data chunk.
  • CRC-32 Checksum: Unrolled the inner checksum loop by 4 bytes.