Per the below graph (from https://github.com/Cyan4973/xxHash), xxH3 is always faster than the older XXH64 and XXH32 functions. About 2-4x, depending on vector instruction set, which is selected automatically at compile time. And nowadays most all servers are AVX-512, so looking at 4x.
Per the below graph (from https://github.com/Cyan4973/xxHash), xxH3 is always faster than the older XXH64 and XXH32 functions. About 2-4x, depending on vector instruction set, which is selected automatically at compile time. And nowadays most all servers are AVX-512, so looking at 4x.
And per these benchmarks https://github.com/rurban/smhasher/ this is the highest throughput of any 32 bit output hash function.
I also forced inlined
XXH_INLINE_ALLit so we can forget about the .c.I left the old
xxhash.handxxhash.cfiles there for now though.