mirror of
https://github.com/fumiama/gofastTEA.git
synced 2026-06-07 17:00:25 +08:00
79 lines
5.0 KiB
Markdown
79 lines
5.0 KiB
Markdown
# gofastTEA
|
||
TEA 编码算法的 PLAN9 汇编优化实现
|
||
|
||
## 编译逻辑
|
||
- 大于等于 1.17 版本且是 amd64，使用非`asm`非内联算法
|
||
- 大于等于 1.17 版本且不是 amd64，使用非`asm`内联算法
|
||
- 小于 1.17 版本且是 amd64，使用`asm`内联算法
|
||
- 小于 1.17 版本且不是 amd64，使用非`asm`内联算法
|
||
|
||
## 1.17 版本及以上
|
||
代码与[MiraiGo](https://github.com/Mrs4s/MiraiGo/blob/574c4e57b1467225f03936342e477ee0d587a2dc/binary/tea.go)相比，替换了随机算法为`runtime.fastrand`，且简化了`Decrypt`，提升速度如下。
|
||
```css
|
||
name old time/op new time/op delta
|
||
TEAen/16-8 241ns ± 1% 225ns ± 1% -6.97% (p=0.000 n=9+10)
|
||
TEAen/256-8 1.71µs ± 1% 1.68µs ± 1% -1.31% (p=0.000 n=10+9)
|
||
TEAen/4K-8 25.0µs ± 1% 25.0µs ± 1% ~ (p=0.497 n=10+9)
|
||
TEAen/32K-8 203µs ± 0% 202µs ± 0% ~ (p=0.079 n=10+9)
|
||
TEAde/16-8 208ns ± 1% 208ns ± 0% ~ (p=0.914 n=9+9)
|
||
TEAde/256-8 1.65µs ± 1% 1.65µs ± 1% ~ (p=0.429 n=9+10)
|
||
TEAde/4K-8 24.7µs ± 1% 24.5µs ± 1% -0.44% (p=0.026 n=9+10)
|
||
TEAde/32K-8 200µs ± 1% 199µs ± 0% -0.37% (p=0.001 n=9+9)
|
||
|
||
name old speed new speed delta
|
||
TEAen/16-8 66.3MB/s ± 1% 71.3MB/s ± 1% +7.50% (p=0.000 n=9+10)
|
||
TEAen/256-8 150MB/s ± 1% 152MB/s ± 1% +1.31% (p=0.000 n=10+9)
|
||
TEAen/4K-8 164MB/s ± 1% 164MB/s ± 1% ~ (p=0.497 n=10+9)
|
||
TEAen/32K-8 162MB/s ± 0% 162MB/s ± 0% ~ (p=0.075 n=10+9)
|
||
TEAde/16-8 154MB/s ± 1% 154MB/s ± 0% ~ (p=0.982 n=9+9)
|
||
TEAde/256-8 165MB/s ± 1% 165MB/s ± 1% ~ (p=0.399 n=9+10)
|
||
TEAde/4K-8 167MB/s ± 1% 168MB/s ± 1% +0.44% (p=0.026 n=9+10)
|
||
TEAde/32K-8 164MB/s ± 1% 165MB/s ± 0% +0.37% (p=0.001 n=9+9)
|
||
```
|
||
|
||
## 1.16 版本及以下
|
||
使用 PLAN9 汇编编写`Encrypt`，内联编写`Decrypt`，替换了加密算法为`runtime.fastrand`，与[MiraiGo](https://github.com/Mrs4s/MiraiGo/blob/574c4e57b1467225f03936342e477ee0d587a2dc/binary/tea.go)代码同在`go1.16`版本下编译相比，提升速度如下（new16.txt）。
|
||
```css
|
||
name old time/op new time/op delta
|
||
TEAen/16-8 252ns ± 0% 227ns ± 0% -10.00% (p=0.000 n=9+10)
|
||
TEAen/256-8 1.77µs ± 1% 1.66µs ± 0% -6.28% (p=0.000 n=9+10)
|
||
TEAen/4K-8 25.9µs ± 0% 24.9µs ± 1% -3.65% (p=0.000 n=10+10)
|
||
TEAen/32K-8 208µs ± 1% 200µs ± 0% -3.70% (p=0.000 n=10+9)
|
||
TEAde/16-8 216ns ± 1% 210ns ± 1% -3.04% (p=0.000 n=10+10)
|
||
TEAde/256-8 1.71µs ± 1% 1.66µs ± 1% -2.93% (p=0.000 n=10+10)
|
||
TEAde/4K-8 25.4µs ± 1% 24.8µs ± 0% -2.36% (p=0.000 n=10+9)
|
||
TEAde/32K-8 206µs ± 0% 200µs ± 0% -2.53% (p=0.000 n=9+9)
|
||
|
||
name old speed new speed delta
|
||
TEAen/16-8 63.5MB/s ± 0% 70.6MB/s ± 0% +11.12% (p=0.000 n=9+10)
|
||
TEAen/256-8 145MB/s ± 1% 154MB/s ± 0% +6.69% (p=0.000 n=9+10)
|
||
TEAen/4K-8 158MB/s ± 0% 164MB/s ± 1% +3.79% (p=0.000 n=10+10)
|
||
TEAen/32K-8 158MB/s ± 1% 164MB/s ± 0% +3.84% (p=0.000 n=10+9)
|
||
TEAde/16-8 148MB/s ± 1% 152MB/s ± 1% +3.12% (p=0.000 n=10+10)
|
||
TEAde/256-8 160MB/s ± 1% 164MB/s ± 1% +3.01% (p=0.000 n=10+10)
|
||
TEAde/4K-8 162MB/s ± 1% 166MB/s ± 0% +2.41% (p=0.000 n=10+9)
|
||
TEAde/32K-8 159MB/s ± 0% 164MB/s ± 0% +2.60% (p=0.000 n=9+9)
|
||
```
|
||
另外[MiraiGo](https://github.com/Mrs4s/MiraiGo/blob/574c4e57b1467225f03936342e477ee0d587a2dc/binary/tea.go)本身在`go1.16`版本与在`go1.17`版本下编译相比，提升速度如下（new17.txt）。
|
||
```css
|
||
name old time/op new time/op delta
|
||
TEAen/16-8 252ns ± 0% 241ns ± 1% -4.09% (p=0.000 n=9+10)
|
||
TEAen/256-8 1.77µs ± 1% 1.70µs ± 0% -3.85% (p=0.000 n=9+10)
|
||
TEAen/4K-8 25.9µs ± 0% 24.9µs ± 1% -3.59% (p=0.000 n=10+10)
|
||
TEAen/32K-8 208µs ± 1% 200µs ± 1% -3.78% (p=0.000 n=10+10)
|
||
TEAde/16-8 216ns ± 1% 208ns ± 1% -3.80% (p=0.000 n=10+10)
|
||
TEAde/256-8 1.71µs ± 1% 1.65µs ± 1% -3.44% (p=0.000 n=10+10)
|
||
TEAde/4K-8 25.4µs ± 1% 24.5µs ± 0% -3.40% (p=0.000 n=10+10)
|
||
TEAde/32K-8 206µs ± 0% 199µs ± 0% -3.36% (p=0.000 n=9+10)
|
||
|
||
name old speed new speed delta
|
||
TEAen/16-8 63.5MB/s ± 0% 66.3MB/s ± 1% +4.27% (p=0.000 n=9+10)
|
||
TEAen/256-8 145MB/s ± 1% 150MB/s ± 0% +4.01% (p=0.000 n=9+10)
|
||
TEAen/4K-8 158MB/s ± 0% 164MB/s ± 1% +3.73% (p=0.000 n=10+10)
|
||
TEAen/32K-8 158MB/s ± 1% 164MB/s ± 1% +3.93% (p=0.000 n=10+10)
|
||
TEAde/16-8 148MB/s ± 1% 154MB/s ± 1% +3.95% (p=0.000 n=10+10)
|
||
TEAde/256-8 160MB/s ± 1% 165MB/s ± 1% +3.55% (p=0.000 n=10+10)
|
||
TEAde/4K-8 162MB/s ± 1% 168MB/s ± 0% +3.52% (p=0.000 n=10+10)
|
||
TEAde/32K-8 159MB/s ± 0% 165MB/s ± 0% +3.45% (p=0.000 n=9+9)
|
||
```
|
||
可见在编码时，在`go1.16`版本下的某些时候（编码大小在`0-16kb`之间），`gofastTEA`比`go1.17`版本的`MiraiGo`实现更快，且整体来看，`gofastTEA`在`go1.16`版本下的执行效率已经可以与`MiraiGo`实现的`go1.17`版本基本持平。 |