Skip to content

v2.2.6

Latest

Choose a tag to compare

@fbernier fbernier released this 09 Sep 19:14
  • Faster short and mid-sized encoding: processes digits in pairs and reuses fixed-width encoding helpers instead of dividing once per digit.
  • Less work for large integers: replaces one full-width quotient calculation with bounded u64 arithmetic.
  • Faster decoding: short inputs avoid wide arithmetic, and byte-indexed lookup tables simplify validation.
  • Smaller decoder-only binaries: encoding tables can now be discarded by the linker. A local linkage probe was about 8 KiB smaller.