I started writing this because all the 64 bit hash functions I came across were either too slow (FNV-1a, one byte at a time processing), or too large spanning hundreds of lines of code, or non-portable due to using hardware specific instructions. Being small and portable, the goal is to be able to use ChibiHash as a good "default" for non-cryptographic 64-bit hashing needs.
Some key features:
Here's some benchmark against other similar hash functions:
Name | Large input (GiB/sec) | Small input (Cycles/Hash) |
---|---|---|
chibihash64 | 18.08 | 49 |
xxhash64 | 12.59 | 50 |
city64 | 14.95 | 35 |
spooky64 | 13.83 | 59 |
It's the fastest of the bunch for large string throughput. For small string (< 32 bytes), cityhash beats it - worth noting that cityhash has hardcoded special cases for input below or equal 32 bytes.
The introduction should make it clear on why you'd want to use this. Here are some reasons to avoid using this:
A list of unofficial ports to other languages is maintained here.