https://linzeyan.github.io/zh-tw/categories/benchmark/Recent content in Benchmark on RickyHugo -- gohugo.iozh-twThu, 02 Sep 2021 12:50:33 +0800https://linzeyan.github.io/zh-tw/posts/2021/20210902-golang-ji-zhun-ce-shi/Thu, 02 Sep 2021 12:50:33 +0800https://linzeyan.github.io/zh-tw/posts/2021/20210902-golang-ji-zhun-ce-shi/<ul> <li><a href="https://mp.weixin.qq.com/s?__biz=MzkxNzAzNDA3Ng==&amp;mid=2247485595&amp;idx=1&amp;sn=9981e6103a6b33cc3a06be0781a2bf10&amp;chksm=c1478da8f63004be44a8c84941280f3888d4e5a2120a4288ed1a21f8a1a82fe65d719c31bae6#rd" target="_blank" rel="noopener">Golang 基准测试</a></li> </ul> <h4 id="基本使用">基本使用</h4> <blockquote> <p>基准测试常用于代码性能测试，函数需要导入 testing 包，并定义以 Benchmark 开头的函数， 参数为 testing.B 指针类型，在测试函数中循环调用函数多次</p></blockquote> <div class="highlight"><pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;"><code class="language-shell" data-lang="shell"><span style="display:flex;"><span>➜ go test -bench<span style="color:#f92672">=</span>. -run<span style="color:#f92672">=</span>none </span></span><span style="display:flex;"><span>goos: darwin </span></span><span style="display:flex;"><span>goarch: amd64 </span></span><span style="display:flex;"><span>pkg: pkg06 </span></span><span style="display:flex;"><span>cpu: Intel<span style="color:#f92672">(</span>R<span style="color:#f92672">)</span> Core<span style="color:#f92672">(</span>TM<span style="color:#f92672">)</span> i7-8850H CPU @ 2.60GHz </span></span><span style="display:flex;"><span>BenchmarkFib-12 <span style="color:#ae81ff">250</span> <span style="color:#ae81ff">4682682</span> ns/op </span></span><span style="display:flex;"><span>PASS </span></span><span style="display:flex;"><span>ok pkg06 1.875s </span></span><span style="display:flex;"><span>➜ go test -bench<span style="color:#f92672">=</span>. -benchmem -run<span style="color:#f92672">=</span>none </span></span><span style="display:flex;"><span>goos: darwin </span></span><span style="display:flex;"><span>goarch: amd64 </span></span><span style="display:flex;"><span>pkg: pkg06 </span></span><span style="display:flex;"><span>cpu: Intel<span style="color:#f92672">(</span>R<span style="color:#f92672">)</span> Core<span style="color:#f92672">(</span>TM<span style="color:#f92672">)</span> i7-8850H CPU @ 2.60GHz </span></span><span style="display:flex;"><span>BenchmarkFib-12 <span style="color:#ae81ff">249</span> <span style="color:#ae81ff">4686452</span> ns/op <span style="color:#ae81ff">0</span> B/op <span style="color:#ae81ff">0</span> allocs/op </span></span><span style="display:flex;"><span>PASS </span></span><span style="display:flex;"><span>ok pkg06 1.854s </span></span></code></pre></div><h4 id="bench-的工作原理">bench 的工作原理</h4> <ul> <li>基准测试函数会被一直调用直到 <code>b.N</code> 无效，它是基准测试循环的次数</li> <li><code>b.N</code> 从 <code>1</code> 开始，如果基准测试函数在 <code>1</code> 秒内就完成 (默认值)，则 <code>b.N</code> 增加，并再次运行基准测试函数</li> <li><code>b.N</code> 的值会按照序列 <code>1,2,5,10,20,50,...</code> 增加，同时再次运行基准测测试函数</li> <li>上述结果解读代表 <code>1</code> 秒内运行了 <code>250</code> 次，每次 <code>4682682 ns</code></li> <li><code>-12</code> 后缀和用于运行次测试的 <code>GOMAXPROCS</code> 值有关。与 <code>GOMAXPROCS</code> 一样，此数字默认为启动时 Go 进程可见的 CPU 数。可以使用 <code>-cpu</code> 标识更改此值，可以传入多个值以列表形式来运行基准测试</li> </ul> <h4 id="传入-cpu-num-进行测试">传入 cpu num 进行测试</h4> <div class="highlight"><pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;"><code class="language-shell" data-lang="shell"><span style="display:flex;"><span>➜ go test -bench<span style="color:#f92672">=</span>. -cpu<span style="color:#f92672">=</span>1,2,4 -benchmem -run<span style="color:#f92672">=</span>none </span></span><span style="display:flex;"><span>goos: darwin </span></span><span style="display:flex;"><span>goarch: amd64 </span></span><span style="display:flex;"><span>pkg: pkg06 </span></span><span style="display:flex;"><span>cpu: Intel<span style="color:#f92672">(</span>R<span style="color:#f92672">)</span> Core<span style="color:#f92672">(</span>TM<span style="color:#f92672">)</span> i7-8850H CPU @ 2.60GHz </span></span><span style="display:flex;"><span>BenchmarkFib <span style="color:#ae81ff">244</span> <span style="color:#ae81ff">4694667</span> ns/op <span style="color:#ae81ff">0</span> B/op <span style="color:#ae81ff">0</span> allocs/op </span></span><span style="display:flex;"><span>BenchmarkFib-2 <span style="color:#ae81ff">255</span> <span style="color:#ae81ff">4721201</span> ns/op <span style="color:#ae81ff">0</span> B/op <span style="color:#ae81ff">0</span> allocs/op </span></span><span style="display:flex;"><span>BenchmarkFib-4 <span style="color:#ae81ff">256</span> <span style="color:#ae81ff">4756392</span> ns/op <span style="color:#ae81ff">0</span> B/op <span style="color:#ae81ff">0</span> allocs/op </span></span><span style="display:flex;"><span>PASS </span></span><span style="display:flex;"><span>ok pkg06 5.826s </span></span></code></pre></div><h4 id="count-多次运行基准测试">count 多次运行基准测试</h4> <blockquote> <p>因为热缩放、内存局部性、后台处理、gc 活动等等会导致单次的误差，所以一般会进行多次测试</p>