# AEGIS for Rust
This is a Rust implementation of the
[AEGIS-128L](https://datatracker.ietf.org/doc/draft-irtf-cfrg-aegis-aead/)
authenticated cipher, ported from the Zig standard library.
AEGIS is extremely fast on CPUs with AES acceleration, has a large nonce size,
and is key committing.
# Cargo flags
- `std`: allow dynamic allocations
`std` is the default.
- `pure-rust`: don't use the `cc` crate to take advantage of the optimized implementation ported from libsodium. Setting this flag will substantially degrade performance. Only required if you need to support old x86_64 or ARM CPUs without AES acceleration.
A benchmark can be run that way:
```sh
export RUSTFLAGS="-C target-cpu=native"
cargo bench
```
For benchmarking, `RUSTFLAGS` is set so that the AES-GCM implementations can take advantage of hardware acceleration.
# Benchmarks
Benchmarks take a 16384 bytes input block. Results are in bytes per second.
## Rust implementations
Crates:
- `aes-gcm`
- `chacha20poly1305`
- `aegis128l`
## Zen2
| aes256-gcm | 934.41 M/s |
| aes128-gcm | 973.18 M/s |
| chacha20-poly1305 | 1.35 G/s |
| aegis128l | 13.70 G/s |
## Zen4
| aes256-gcm | 1.36 G/s |
| aes128-gcm | 1.43 G/s |
| chacha20-poly1305 | 1.82 G/s |
| aegis128l | 15.43 G/s |
## Macbook Pro - Apple M1
| aes256-gcm | 139.66 M/s |
| aes128-gcm | 173.09 M/s |
| chacha20-poly1305 | 265.48 M/s |
| aegis128l | 13.88 G/s |
## WebAssembly (Wasmtime, Zen4)
| aes256-gcm | 62.97 M/s |
| aes128-gcm | 73.83 M/s |
| chacha20-poly1305 | 88.92 M/s |
| aegis128l | 537.49 M/s |
## WebAssembly (Wasmtime, Apple M1)
| aes256-gcm | 49.43 M/s |
| aes128-gcm | 59.37 M/s |
| chacha20-poly1305 | 177.85 M/s |
| aegis128l | 533.85 M/s |
## Other implementations, Zen2
| aes256-gcm (OpenSSL) | 4.97 G/s |
| aes128-gcm (OpenSSL) | 6.89 G/s |
| chacha20-poly1305 (OpenSSL) | 2.67 G/s |
| aes128-ocb (OpenSSL) | 7.10 G/s |
| aegis128l (Zig) | 14.08 G/s |