blake3

go get lukechampine.com/blake3

blake3 implements the BLAKE3 cryptographic hash function. This implementation aims to be performant without sacrificing (too much) readability, in the hopes of eventually landing in x/crypto.

The pure-Go code is fairly well-optimized, achieving throughput of ~600 MB/s. There is a separate code path for small inputs (up to 64 bytes) that runs in ~100 ns. On CPUs with AVX2 support, larger inputs (>=2 KB) are handled by an avo-generated assembly routine that compresses 8 chunks in parallel, achieving throughput of ~2600 MB/s. Once AVX-512 support is added to avo, it will be possible to compress 16 chunks in parallel, which should roughly double throughput for sufficiently large inputs.

Contributions are greatly appreciated. All contributors are eligible to receive an Urbit planet.

Benchmarks

Tested on an i5-7600K @ 3.80GHz.

BenchmarkSum256/64           105 ns/op       609.51 MB/s
BenchmarkSum256/1024        1778 ns/op       576.00 MB/s
BenchmarkSum256/65536      24785 ns/op      2644.15 MB/s
BenchmarkWrite               389 ns/op      2631.78 MB/s
BenchmarkXOF                1591 ns/op       643.80 MB/s