golang逃逸分析和競爭檢測

最近在線上發現一塊代碼邏輯在執行N次耗時波動很大1ms~800ms,最開始以為是gc的問題，對代碼進行逃逸分析，看哪些變量被分配到堆上了，后來發現是並發編程時對一個切片並發的寫,導致存在競爭,類似下面的代碼

func main() {
	//var count int
	array := make([]int, 100000)
	wg := new(sync.WaitGroup)
	for i := 0; i < 10; i++ {
		wg.Add(1)
		go func(a []int) {
			now := time.Now()
			print(a)
			fmt.Println("耗時:", time.Since(now))
			wg.Done()
		}(array)
	}
	wg.Wait()
}

func print(array []int) {
	array[0] = 1
	array[1] = 1
	for i := 0; i < len(array); i++ {

	}
	//fmt.Println(array)
}

output:

耗時: 85.532µs
耗時: 49.543µs
耗時: 53.306µs
耗時: 53.365µs
耗時: 47.73µs
耗時: 48.098µs
耗時: 70.815µs
耗時: 71.15µs
耗時: 89.213µs
耗時: 60.797µs

首先試一試逃逸分析:

go build -gcflags '-m -l' main.go 

./main.go:27:20: print array does not escape
./main.go:11:15: make([]int, 100000) escapes to heap
./main.go:12:11: new(sync.WaitGroup) escapes to heap
./main.go:15:13: make([]int, 100000) escapes to heap
./main.go:17:6: func literal escapes to heap
./main.go:17:6: func literal escapes to heap
./main.go:20:16: "耗時:" escapes to heap
./main.go:20:37: time.Since(now) escapes to heap
./main.go:21:4: leaking closure reference wg
./main.go:17:15: main.func1 a does not escape
./main.go:20:15: main.func1 ... argument does not escape

結論:切片array由於size太大了被分配到堆上了，字符串"耗時:"被分配到堆上，這里分配到堆上的變量被頻繁創建地有newA和字符串"耗時:"，newA可以采用變量池sync.Pool解決，字符串應該寫成常量形式

耗時47us~89us，很不穩定，對其進行競爭檢測

運行命令

go run -race main.go

output:

==================
WARNING: DATA RACE
Write at 0x00c420092000 by goroutine 7:
  main.print()
      /Users/qianxiaoli/work/mygopath/src/testpprof/bug/main.go:26 +0x49
  main.main.func1()
      /Users/qianxiaoli/work/mygopath/src/testpprof/bug/main.go:17 +0x8c

Previous write at 0x00c420092000 by goroutine 6:
  main.print()
      /Users/qianxiaoli/work/mygopath/src/testpprof/bug/main.go:26 +0x49
  main.main.func1()
      /Users/qianxiaoli/work/mygopath/src/testpprof/bug/main.go:17 +0x8c

Goroutine 7 (running) created at:
  main.main()
      /Users/qianxiaoli/work/mygopath/src/testpprof/bug/main.go:15 +0xe3

Goroutine 6 (running) created at:
  main.main()
      /Users/qianxiaoli/work/mygopath/src/testpprof/bug/main.go:15 +0xe3
==================
耗時: 58.625µs
==================
WARNING: DATA RACE
Write at 0x00c420092008 by goroutine 7:
  main.print()
      /Users/qianxiaoli/work/mygopath/src/testpprof/bug/main.go:27 +0x6d
  main.main.func1()
      /Users/qianxiaoli/work/mygopath/src/testpprof/bug/main.go:17 +0x8c

Previous write at 0x00c420092008 by goroutine 6:
  main.print()
      /Users/qianxiaoli/work/mygopath/src/testpprof/bug/main.go:27 +0x6d
  main.main.func1()
      /Users/qianxiaoli/work/mygopath/src/testpprof/bug/main.go:17 +0x8c

Goroutine 7 (running) created at:
  main.main()
      /Users/qianxiaoli/work/mygopath/src/testpprof/bug/main.go:15 +0xe3

Goroutine 6 (running) created at:
  main.main()
      /Users/qianxiaoli/work/mygopath/src/testpprof/bug/main.go:15 +0xe3
==================

結論：
顯然並發代碼存在競爭，print函數對array並發寫操作導致競爭，執行耗時變長。由於切片在拷貝時，底層的數組還是同一個，所以並發寫同一個數組會產生競爭。

將代碼加入變量池及切片拷貝

var arrayPool = sync.Pool{
	New: func() interface{} {
		a := make([]int, 2)
		return a
	},
}

func main() {
	array := make([]int, 2)
	wg := new(sync.WaitGroup)
	for i := 0; i < 10000; i++ {
		wg.Add(1)
		newA := arrayPool.Get().([]int)
		copy(newA, array)
		go func(a []int) {
			now := time.Now()
			print(a)
			fmt.Println("耗時:", time.Since(now))
			wg.Done()
		}(newA)
	}
	wg.Wait()
}

func print(array []int) {
	array[0] = 1
	array[1] = 1
	for i := 0; i < len(array); i++ {

	}
	arrayPool.Put(array)
	//fmt.Println(array)
}