Codebase list golang-github-minio-sha256-simd / c985c18
Update README to reflect AVX2 support for crypto/sha256 frankw 6 years ago
1 changed file(s) with 2 addition(s) and 18 deletion(s). Raw diff Collapse all Expand all
+2
-18
README.md less more
00 # sha256-simd
11
22 Accelerate SHA256 computations in pure Go for both Intel (AVX2, AVX, SSE) as well as ARM (arm64) platforms.
3
4 Update: As of Go 1.8, `crypto/sha256` offers similar performance for AVX2.
35
46 ## Introduction
57
4547 | 2.4 GHz Intel Xeon CPU E5-2620 v3 | minio/sha256-simd (AVX2) | 355.0 MB/s | 1.88x |
4648 | 2.4 GHz Intel Xeon CPU E5-2620 v3 | minio/sha256-simd (AVX) | 306.0 MB/s | 1.62x |
4749 | 2.4 GHz Intel Xeon CPU E5-2620 v3 | minio/sha256-simd (SSE) | 298.7 MB/s | 1.58x |
48 | 2.4 GHz Intel Xeon CPU E5-2620 v3 | crypto/sha256 | 189.2 MB/s | |
4950 | 1.2 GHz ARM Cortex-A53 | crypto/sha256 | 6.1 MB/s | |
5051
5152 Note that the AVX2 version is measured with the "unrolled"/"demacro-ed" version. Due to some Golang assembly restrictions the AVX2 version that uses `defines` loses about 15% performance (you can see the macrofied version, which is a little bit easier to read, [here](https://github.com/minio/sha256-simd/blob/e1b0a493b71bb31e3f1bf82d3b8cbd0d6960dfa6/sha256blockAvx2_amd64.s)).
118119
119120 Example performance metrics were generated on Intel(R) Xeon(R) CPU E5-2620 v3 @ 2.40GHz - 6 physical cores, 12 logical cores running Ubuntu GNU/Linux with kernel version 4.4.0-24-generic (vanilla with no optimizations).
120121
121 ### AVX2
122
123 ```
124 $ benchcmp go.txt avx2.txt
125 benchmark old ns/op new ns/op delta
126 BenchmarkHash8Bytes-12 446 364 -18.39%
127 BenchmarkHash1K-12 5919 3279 -44.60%
128 BenchmarkHash8K-12 43791 23655 -45.98%
129 BenchmarkHash1M-12 5544989 2969305 -46.45%
130
131 benchmark old MB/s new MB/s speedup
132 BenchmarkHash8Bytes-12 17.93 21.96 1.22x
133 BenchmarkHash1K-12 172.98 312.27 1.81x
134 BenchmarkHash8K-12 187.07 346.31 1.85x
135 BenchmarkHash1M-12 189.10 353.14 1.87x
136 ```
137
138122 ### AVX
139123
140124 ```