Tillich-Zémor hashing golang implementation
Find a file
Evgenii Stratonikov 3de3046074 tz: optimize AVX2 implementation
1. Perform masking with 2 instructions instead of 3 (use arithmetic
   shift).
2. Broadcast data byte in one instruction at the start of byte-processing
3. Reorder instructions to reduce the amount of data hazards and resources
   contention.

```
name               old time/op    new time/op    delta
Sum/AVX2_digest-8    1.39ms ± 0%    1.22ms ± 0%  -12.18%  (p=0.000 n=9+7)

name               old speed      new speed      delta
Sum/AVX2_digest-8  71.7MB/s ± 0%  81.7MB/s ± 0%  +13.87%  (p=0.000 n=9+7)
```

Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
2022-03-22 12:25:13 +03:00
.github/workflows go.mod: bump to go1.16 2022-03-21 12:30:08 +03:00
cmd tz: use build tags for different implemenations 2022-03-21 12:30:08 +03:00
gf127 *: rename ByteArray to Bytes 2022-03-21 12:30:08 +03:00
tz tz: optimize AVX2 implementation 2022-03-22 12:25:13 +03:00
.gitignore Ignore vendor and binary 2020-01-16 11:30:45 +03:00
auto.sh Initial 2018-12-29 16:04:17 +03:00
benchmark benchmark: fix shellcheck issues 2022-03-21 12:30:08 +03:00
Dockerfile Update alpine image, fixup for Makefile, fixup for benchmark 2020-01-16 11:30:46 +03:00
go.mod go.mod: bump to go1.16 2022-03-21 12:30:08 +03:00
go.sum go.mod: update dependencies 2022-01-24 13:58:13 +03:00
LICENSE Initial 2018-12-29 16:04:17 +03:00
Makefile Makefile: add target for testing generic implementation 2022-03-21 12:30:08 +03:00
README.md Update benchmark result in README.md 2019-10-16 15:11:57 +03:00

Demo

asciicast

In project root:

# show help
make
# run auto demo
make auto

Homomorphic hashing in golang

Package tz containts pure-Go implementation of hashing function described by Tillich and Źemor in [1] .

There are existing implementations already (e.g. [2]), however they are written in C.

Package gf127 contains arithmetic in GF(2^127) with x^127+x^63+1 as reduction polynomial.

Description

It can be used instead of Merkle-tree for data-validation, because homomorphic hashes are concatenable: hash sum of data can be calculated based on hashes of chunks.

The example of how it works can be seen in tests.

Benchmarks

go vs AVX vs AVX2 version

BenchmarkSum/AVX_digest-8             308       3889484 ns/op          25.71 MB/s         5 allocs/op
BenchmarkSum/AVXInline_digest-8       457       2455437 ns/op          40.73 MB/s         5 allocs/op
BenchmarkSum/AVX2_digest-8            399       3031102 ns/op          32.99 MB/s         3 allocs/op
BenchmarkSum/AVX2Inline_digest-8      602       2077719 ns/op          48.13 MB/s         3 allocs/op
BenchmarkSum/PureGo_digest-8           68       17795480 ns/op          5.62 MB/s         5 allocs/op

Contributing

At this moment, we do not accept contributions. Follow us.

Makefile

→ make
  Usage:

    make <target>

  Targets:

    attach   Attach to existing container
    auto     Auto Tillich-Zémor hasher demo
    down     Stop demo container
    help     Show this help prompt
    up       Run Tillich-Zémor hasher demo

Links

[1] https://link.springer.com/content/pdf/10.1007/3-540-48658-5_5.pdf

[2] https://github.com/srijs/hwsl2-core