README
¶
( temporary fork, waiting for upstream hash.Hash interface feature add ) ( PR request open, see github.com/zeebo/blake3/pull/25 )
BLAKE3
Pure Go implementation of BLAKE3 with AVX2 and SSE4.1 acceleration.
Special thanks to the excellent avo making writing vectorized version much easier.
Benchmarks
Caveats
This library makes some different design decisions than the upstream Rust crate around internal buffering. Specifically, because it does not target the embedded system space, nor does it support multithreading, it elects to do its own internal buffering. This means that a user does not have to worry about providing large enough buffers to get the best possible performance, but it does worse on smaller input sizes. So some notes:
- The Rust benchmarks below are all single-threaded to match this Go implementation.
- I make no attempt to get precise measurements (cpu throttling, noisy environment, etc.) so please benchmark on your own systems.
- These benchmarks are run on an i7-6700K which does not support AVX-512, so Rust is limited to use AVX2 at sizes above 8 kib.
- I tried my best to make them benchmark the same thing, but who knows? 😄
Charts
In this case, both libraries are able to avoid a lot of data copying and will use vectorized instructions to hash as fast as possible, and perform similarly.
For incremental writes, you must provide the Rust version large enough buffers so that it can use vectorized instructions. This Go library performs consistently regardless of the size being sent into the update function.
The downside of internal buffering is most apparent with small sizes as most time is spent initializing the hasher state. In terms of hashing rate, the difference is 3-4x, but in an absolute sense it's ~100ns (see tables below). If you wish to hash a large number of very small strings and you care about those nanoseconds, be sure to use the Reset method to avoid re-initializing the state.
Timing Tables
Small
| Size | Full Buffer | Reset | Full Buffer Rate | Reset Rate | |
|---|---|---|---|---|---|
| 64 b | 205ns |
86.5ns |
312MB/s |
740MB/s |
|
| 256 b | 364ns |
250ns |
703MB/s |
1.03GB/s |
|
| 512 b | 575ns |
468ns |
892MB/s |
1.10GB/s |
|
| 768 b | 795ns |
682ns |
967MB/s |
1.13GB/s |
Large
| Size | Incremental | Full Buffer | Reset | Incremental Rate | Full Buffer Rate | Reset Rate | |
|---|---|---|---|---|---|---|---|
| 1 kib | 1.02µs |
1.01µs |
891ns |
1.00GB/s |
1.01GB/s |
1.15GB/s |
|
| 2 kib | 2.11µs |
2.07µs |
1.95µs |
968MB/s |
990MB/s |
1.05GB/s |
|
| 4 kib | 2.28µs |
2.15µs |
2.05µs |
1.80GB/s |
1.90GB/s |
2.00GB/s |
|
| 8 kib | 2.64µs |
2.52µs |
2.44µs |
3.11GB/s |
3.25GB/s |
3.36GB/s |
|
| 16 kib | 4.93µs |
4.54µs |
4.48µs |
3.33GB/s |
3.61GB/s |
3.66GB/s |
|
| 32 kib | 9.41µs |
8.62µs |
8.54µs |
3.48GB/s |
3.80GB/s |
3.84GB/s |
|
| 64 kib | 18.2µs |
16.7µs |
16.6µs |
3.59GB/s |
3.91GB/s |
3.94GB/s |
|
| 128 kib | 36.3µs |
32.9µs |
33.1µs |
3.61GB/s |
3.99GB/s |
3.96GB/s |
|
| 256 kib | 72.5µs |
65.7µs |
66.0µs |
3.62GB/s |
3.99GB/s |
3.97GB/s |
|
| 512 kib | 145µs |
131µs |
132µs |
3.60GB/s |
4.00GB/s |
3.97GB/s |
|
| 1024 kib | 290µs |
262µs |
262µs |
3.62GB/s |
4.00GB/s |
4.00GB/s |
No ASM
| Size | Incremental | Full Buffer | Reset | Incremental Rate | Full Buffer Rate | Reset Rate | |
|---|---|---|---|---|---|---|---|
| 64 b | 253ns |
254ns |
134ns |
253MB/s |
252MB/s |
478MB/s |
|
| 256 b | 553ns |
557ns |
441ns |
463MB/s |
459MB/s |
580MB/s |
|
| 512 b | 948ns |
953ns |
841ns |
540MB/s |
538MB/s |
609MB/s |
|
| 768 b | 1.38µs |
1.40µs |
1.35µs |
558MB/s |
547MB/s |
570MB/s |
|
| 1 kib | 1.77µs |
1.77µs |
1.70µs |
577MB/s |
580MB/s |
602MB/s |
|
| 1024 kib | 880µs |
883µs |
878µs |
596MB/s |
595MB/s |
598MB/s |
The speed caps out at around 1 kib, so most rows have been elided from the presentation.
Documentation
¶
Overview ¶
Package blake3 provides an SSE4.1/AVX2 accelerated BLAKE3 implementation.
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func DeriveKey ¶
DeriveKey derives a key based on reusable key material of any length, in the given context. The key will be stored in out, using all of its current length.
Context strings must be hardcoded constants, and the recommended format is "[application] [commit timestamp] [purpose]", e.g., "example.com 2019-12-25 16:18:03 session tokens v1".
Types ¶
type Digest ¶
type Digest struct {
// contains filtered or unexported fields
}
Digest captures the state of a Hasher allowing reading and seeking through the output stream.
type Hasher ¶
type Hasher struct {
// contains filtered or unexported fields
}
Hasher is a hash.Hash for BLAKE3.
func New ¶
func New() *Hasher
New returns a new Hasher that has a digest size of 32 bytes.
If you need more or less output bytes than that, use Digest method.
func New512 ¶
func New512() *Hasher
New512 returns a new Hasher that has a digest size of 64 byties.
func NewDeriveKey ¶
NewDeriveKey returns a Hasher that is initialized with the context string. See DeriveKey for details. It has a digest size of 32 bytes.
If you need more or less output bytes than that, use the Digest method.
func NewKeyed ¶
NewKeyed returns a new Hasher that uses the 32 byte input key and has a digest size of 32 bytes.
If you need more or less output bytes than that, use the Digest method.
func (*Hasher) BlockSize ¶
BlockSize implements part of the hash.Hash interface. It returns the most natural size to write to the Hasher.
func (*Hasher) Clone ¶
Clone returns a new Hasher with the same internal state.
Modifying the resulting Hasher will not modify the original Hasher, and vice versa.
func (*Hasher) Digest ¶
Digest takes a snapshot of the hash state and returns an object that can be used to read and seek through 2^64 bytes of digest output.
func (*Hasher) Reset ¶
func (h *Hasher) Reset()
Reset implements part of the hash.Hash interface. It causes the Hasher to act as if it was newly created.
func (*Hasher) Size ¶
Size implements part of the hash.Hash interface. It returns the number of bytes the hash will output in Sum.
func (*Hasher) Sum ¶
Sum implements part of the hash.Hash interface. It appends the digest of the Hasher to the provided buffer and returns it.
Source Files
Directories