komihash (rust)

Komihash is a hash function passes all [smhasher] tests with extremely high performance. A PRNG named komirand is also provided that passes PractRand test. See [the original C implementation] for more details.

Performance

This crate archives almost the same performance with the original C implementation. The benchmark is done on an Apple Macbook Pro (Apple M1 Pro).

``` test benchsimplewith00000bytesx10input ... bench: 17 ns/iter (+/- 0) test benchsimplewith00001bytesx10input ... bench: 26 ns/iter (+/- 0) test benchsimplewith00002bytesx10input ... bench: 27 ns/iter (+/- 1) test benchsimplewith00003bytesx10input ... bench: 28 ns/iter (+/- 1) test benchsimplewith00004bytesx10input ... bench: 25 ns/iter (+/- 0) test benchsimplewith00005bytesx10input ... bench: 26 ns/iter (+/- 0) test benchsimplewith00006bytesx10input ... bench: 28 ns/iter (+/- 1) test benchsimplewith00007bytesx10input ... bench: 29 ns/iter (+/- 2) test benchsimplewith00008bytesx10input ... bench: 21 ns/iter (+/- 1) test benchsimplewith00009bytesx10input ... bench: 27 ns/iter (+/- 1) test benchsimplewith00010bytesx10input ... bench: 29 ns/iter (+/- 0) test benchsimplewith00011bytesx10input ... bench: 30 ns/iter (+/- 1) test benchsimplewith00012bytesx10input ... bench: 27 ns/iter (+/- 1) test benchsimplewith00013bytesx10input ... bench: 29 ns/iter (+/- 2) test benchsimplewith00014bytesx10input ... bench: 29 ns/iter (+/- 0) test benchsimplewith00015bytesx10input ... bench: 29 ns/iter (+/- 1) test benchsimplewith00016bytesx10input ... bench: 26 ns/iter (+/- 1) test benchsimplewith00024bytesx10input ... bench: 27 ns/iter (+/- 1) test benchsimplewith00032bytesx10input ... bench: 33 ns/iter (+/- 1) test benchsimplewith00048bytesx10input ... bench: 42 ns/iter (+/- 3) test benchsimplewith00064bytesx10input ... bench: 58 ns/iter (+/- 3) test benchsimplewith00096bytesx10input ... bench: 70 ns/iter (+/- 4) test benchsimplewith00128bytesx10input ... bench: 79 ns/iter (+/- 2) test benchsimplewith00192bytesx10input ... bench: 101 ns/iter (+/- 5) test benchsimplewith00256bytesx10input ... bench: 124 ns/iter (+/- 7) test benchsimplewith01024bytesx10input ... bench: 445 ns/iter (+/- 18) test benchsimplewith04096bytesx10input ... bench: 1,760 ns/iter (+/- 125) test benchsimplewith16384bytesx10input ... bench: 7,036 ns/iter (+/- 445) test benchsimplewith65536bytesx10_input ... bench: 28,130 ns/iter (+/- 944)

test result: ok. 0 passed; 0 failed; 0 ignored; 29 measured; 0 filtered out; finished in 21.00s

```

or in a more readable format

| Input size (Byte) | Time (ns) | Throughput(GB/s) | |---------------------|-----------|--------------------| | 0 | 17 | N/A | | 1 | 26 | 0.04 | | 2 | 27 | 0.07 | | 3 | 28 | 0.11 | | 4 | 25 | 0.16 | | 5 | 26 | 0.19 | | 6 | 28 | 0.21 | | 7 | 29 | 0.24 | | 8 | 21 | 0.38 | | 9 | 27 | 0.33 | | 10 | 29 | 0.34 | | 11 | 30 | 0.37 | | 12 | 27 | 0.44 | | 13 | 29 | 0.45 | | 14 | 29 | 0.48 | | 15 | 29 | 0.51 | | 16 | 26 | 0.61 | | 24 | 27 | 0.89 | | 32 | 33 | 0.97 | | 48 | 42 | 1.14 | | 64 | 58 | 1.10 | | 96 | 70 | 1.37 | | 128 | 79 | 1.62 | | 192 | 101 | 1.90 | | 256 | 124 | 2.06 | | 1024 | 445 | 2.29 | | 4096 | 1760 | 2.33 | | 16384 | 7036 | 2.34 | | 65536 | 28130 | 2.33 |

License

MIT