以斐波那契数列作为测试用例:

package fib

func Fib(n int) int {
	switch n {
	case 0:
		return 0
	case 1:
		return 1
	case 2:
		return 2
	default:
		return Fib(n-1) + Fib(n-2)
	}
}
Benchmarkstesting*testing.B*testing.T
方法 说明
Log 打印日志，同时结束测试
Logf 格式化打印日志，同时结束测试
Error 打印错误日志，同时结束测试
Errorf 格式化打印错误日志，同时结束测试
Fatal 打印致命日志，同时结束测试
Fatalf 格式化打印致命日志，同时结束测试 
testing.B
fib_test.go
package fib

import (
	"testing"
)

func BenchmarkFib20(b *testing.B) {
	for n := 0; n < b.N; n++ {
		Fib(20) // run the Fib function b.N times
	}
}
go test -bench=.

goos: linux
goarch: amd64
BenchmarkFib20-8 50000 37932 ns/op
PASS

Windows 下使用 go test 命令行时，-bench=.应写为-bench="."
-cpuGOMAXPROCS-cpu=1,2,4

$ go test -bench=. -cpu=1,2,4
goos: linux
goarch: amd64
BenchmarkFib20 50000 36877 ns/op
BenchmarkFib20-2 50000 36843 ns/op
BenchmarkFib20-4 30000 36231 ns/op

-benchtime

$ go test -bench=. -benchtime=3s
goos: linux
goarch: amd64
BenchmarkFib20-8 100000 46070 ns/op
PASS
ok _/home/shikanon/study/fib 5.018s

你会发现总的运行时间并不是3s，而且不同的运行其时间会有差异，这主要是因为由于GC活动、后台运行程序、内存位置、CPU的调整频率所影响。

-count

$ go test -bench=. -count=10
goos: linux
goarch: amd64
BenchmarkFib20-8 50000 38395 ns/op
BenchmarkFib20-8 50000 37019 ns/op
BenchmarkFib20-8 50000 36974 ns/op
BenchmarkFib20-8 50000 41050 ns/op
BenchmarkFib20-8 50000 36131 ns/op
BenchmarkFib20-8 30000 37317 ns/op
BenchmarkFib20-8 30000 42872 ns/op
BenchmarkFib20-8 30000 41711 ns/op
BenchmarkFib20-8 50000 36517 ns/op
BenchmarkFib20-8 50000 40240 ns/op
PASS
ok _/home/shikanon/study/fib 20.996s

benchstatbenchstatpref

$benchstat old.txt
name time/op
Fib20-8 39.4µs ±13%

prefgolang/xgit clonebenchstatgo build -o $GOROOT/bin/benchstat main.go
go test -c

./fib2.test -bench=. -test.count=10 > old.txt
./fib.test -bench=. -test.count=10 > new.txt
$ benchstat old.txt new.txt
name old time/op new time/op delta
Fib20-8 44.3µs ± 6% 25.6µs ± 2% -42.31% (p=0.000 n=10+10)

关于 golang 的 for range 的性能比较

for range
package main

import (
	"testing"
)

func BenchmarkArray(b *testing.B) {
	var a = 0
	for i := 0; i < b.N; i++ {
		var times [100][0]int
		for range times {
			a += 1
		}
	}
}

func BenchmarkInt(b *testing.B) {
	var a = 0
	for i := 0; i < b.N; i++ {
		for i := 0; i < 100; i++ {
			a += 1
		}
	}
}
[100][0]int[0]int

测试结果：

go test -bench=.
goos: windows
goarch: amd64
BenchmarkArray-8 50000000 36.1 ns/op
BenchmarkInt-8 20000000 65.9 ns/op
PASS

for rangefor

切片高效删除指定元素操作

利用了零长切片的特性，实现高效删除技巧：

func TrimSpaceWithAssignment(s []byte) []byte {
	var b = make([]byte, len(s))
	for i, x := range s {
		if x != ' ' {
			b[i] = x
		}
	}
	return b
}

func TrimSpaceWithNil(s []byte) []byte {
	var b = make([]byte, 0, len(s))
	for _, x := range s {
		if x != ' ' {
			b = append(b, x)
		}
	}
	return b
}

func TrimSpaceWithOrigin(s []byte) []byte {
	b := s[:0]
	for _, x := range s {
		if x != ' ' {
			b = append(b, x)
		}
	}
	return b
}

func BenchmarkAssignment(b *testing.B) {
	s := []byte("asdf asdf sa fd dsa f d")
	for i := 0; i < b.N; i++ {
		TrimSpaceWithAssignment(s)
	}
}

func BenchmarkNilSlice(b *testing.B) {
	s := []byte("asdf asdf sa fd dsa f d")
	for i := 0; i < b.N; i++ {
		TrimSpaceWithNil(s)
	}
}

func BenchmarkOrignSlice(b *testing.B) {
	s := []byte("asdf asdf sa fd dsa f d")
	for i := 0; i < b.N; i++ {
		TrimSpaceWithOrigin(s)
	}
}

测试结果：

go test -bench=.
goos: windows
goarch: amd64
BenchmarkAssignment-8 30000000 43.7 ns/op
BenchmarkNilSlice-8 30000000 47.6 ns/op
BenchmarkOrignSlice-8 50000000 22.7 ns/op

参考