Go 程序代码分析（Profiling）

$ go version
go version devel +08d20469cc20 Tue Mar 26 08:27:18 2013 +0100 linux/amd64
$ g++ --version
g++ (GCC) 4.8.0
Copyright (C) 2013 Free Software Foundation, Inc.
...
$Copy

$ sudo bash
# for i in /sys/devices/system/cpu/cpu[0-7]
do
    echo performance > $i/cpufreq/scaling_governor
done
#Copy

$ cat xtime
#!/bin/sh
/usr/bin/time -f '%Uu %Ss %er %MkB %C' "$@"
$

$ make havlak1cc
g++ -O3 -o havlak1cc havlak1.cc
$ ./xtime ./havlak1cc
# of loops: 76002 (total 3800100)
loop-0, nest: 0, depth: 0
17.70u 0.05s 17.80r 715472kB ./havlak1cc
$

$ make havlak1
go build havlak1.go
$ ./xtime ./havlak1
# of loops: 76000 (including 1 artificial root node)
25.05u 0.11s 25.20r 1334032kB ./havlak1
$Copy

var cpuprofile = flag.String("cpuprofile", "", "write cpu profile to file")

func main() {
    flag.Parse()
    if *cpuprofile != "" {
        f, err := os.Create(*cpuprofile)
        if err != nil {
            log.Fatal(err)
        }
        pprof.StartCPUProfile(f)
        defer pprof.StopCPUProfile()
    }
    ...Copy

$ make havlak1.prof
./havlak1 -cpuprofile=havlak1.prof
# of loops: 76000 (including 1 artificial root node)
$ go tool pprof havlak1 havlak1.prof
Welcome to pprof!  For help, type 'help'.
(pprof)Copy

(pprof) top10
Total: 2525 samples
     298  11.8%  11.8%      345  13.7% runtime.mapaccess1_fast64
     268  10.6%  22.4%     2124  84.1% main.FindLoops
     251   9.9%  32.4%      451  17.9% scanblock
     178   7.0%  39.4%      351  13.9% hash_insert
     131   5.2%  44.6%      158   6.3% sweepspan
     119   4.7%  49.3%      350  13.9% main.DFS
      96   3.8%  53.1%       98   3.9% flushptrbuf
      95   3.8%  56.9%       95   3.8% runtime.aeshash64
      95   3.8%  60.6%      101   4.0% runtime.settype_flush
      88   3.5%  64.1%      988  39.1% runtime.mallocgcCopy

(pprof) top5 -cum
Total: 2525 samples
       0   0.0%   0.0%     2144  84.9% gosched0
       0   0.0%   0.0%     2144  84.9% main.main
       0   0.0%   0.0%     2144  84.9% runtime.main
       0   0.0%   0.0%     2124  84.1% main.FindHavlakLoops
     268  10.6%  10.6%     2124  84.1% main.FindLoops
(pprof) top5 -cumCopy

(pprof) webCopy

(pprof) web mapaccess1Copy

(pprof) list DFS
Total: 2525 samples
ROUTINE ====================== main.DFS in /home/rsc/g/benchgraffiti/havlak/havlak1.go
   119    697 Total samples (flat / cumulative)
     3      3  240: func DFS(currentNode *BasicBlock, nodes []*UnionFindNode, number map[*BasicBlock]int, last []int, current int) int {
     1      1  241:     nodes[current].Init(currentNode, current)
     1     37  242:     number[currentNode] = current
     .      .  243:
     1      1  244:     lastid := current
    89     89  245:     for _, target := range currentNode.OutEdges {
     9    152  246:             if number[target] == unvisited {
     7    354  247:                     lastid = DFS(target, nodes, number, last, lastid+1)
     .      .  248:             }
     .      .  249:     }
     7     59  250:     last[number[currentNode]] = lastid
     1      1  251:     return lastid
(pprof)Copy

$ make havlak2
go build havlak2.go
$ ./xtime ./havlak2
# of loops: 76000 (including 1 artificial root node)
16.55u 0.11s 16.69r 1321008kB ./havlak2
$Copy

$ make havlak2.prof
./havlak2 -cpuprofile=havlak2.prof
# of loops: 76000 (including 1 artificial root node)
$ go tool pprof havlak2 havlak2.prof
Welcome to pprof!  For help, type 'help'.
(pprof)
(pprof) top5
Total: 1652 samples
     197  11.9%  11.9%      382  23.1% scanblock
     189  11.4%  23.4%     1549  93.8% main.FindLoops
     130   7.9%  31.2%      152   9.2% sweepspan
     104   6.3%  37.5%      896  54.2% runtime.mallocgc
      98   5.9%  43.5%      100   6.1% flushptrbuf
(pprof)Copy

var memprofile = flag.String("memprofile", "", "write memory profile to this file")
...

    FindHavlakLoops(cfgraph, lsgraph)
    if *memprofile != "" {
        f, err := os.Create(*memprofile)
        if err != nil {
            log.Fatal(err)
        }
        pprof.WriteHeapProfile(f)
        f.Close()
        return
    }Copy

$ make havlak3.mprof
go build havlak3.go
./havlak3 -memprofile=havlak3.mprof
$Copy

$ go tool pprof havlak3 havlak3.mprof
Adjusting heap profiles for 1-in-524288 sampling rate
Welcome to pprof!  For help, type 'help'.
(pprof) top5
Total: 82.4 MB
    56.3  68.4%  68.4%     56.3  68.4% main.FindLoops
    17.6  21.3%  89.7%     17.6  21.3% main.(*CFG).CreateNode
     8.0   9.7%  99.4%     25.6  31.0% main.NewBasicBlockEdge
     0.5   0.6% 100.0%      0.5   0.6% itab
     0.0   0.0% 100.0%      0.5   0.6% fmt.init
(pprof)Copy

(pprof) list FindLoops
Total: 82.4 MB
ROUTINE ====================== main.FindLoops in /home/rsc/g/benchgraffiti/havlak/havlak3.go
  56.3   56.3 Total MB (flat / cumulative)
...
   1.9    1.9  268:     nonBackPreds := make([]map[int]bool, size)
   5.8    5.8  269:     backPreds := make([][]int, size)
     .      .  270:
   1.9    1.9  271:     number := make([]int, size)
   1.9    1.9  272:     header := make([]int, size, size)
   1.9    1.9  273:     types := make([]int, size, size)
   1.9    1.9  274:     last := make([]int, size, size)
   1.9    1.9  275:     nodes := make([]*UnionFindNode, size, size)
     .      .  276:
     .      .  277:     for i := 0; i < size; i++ {
   9.5    9.5  278:             nodes[i] = new(UnionFindNode)
     .      .  279:     }
...
     .      .  286:     for i, bb := range cfgraph.Blocks {
     .      .  287:             number[bb.Name] = unvisited
  29.5   29.5  288:             nonBackPreds[i] = make(map[int]bool)
     .      .  289:     }
...Copy

$ go tool pprof --inuse_objects havlak3 havlak3.mprof
Adjusting heap profiles for 1-in-524288 sampling rate
Welcome to pprof!  For help, type 'help'.
(pprof) list FindLoops
Total: 1763108 objects
ROUTINE ====================== main.FindLoops in /home/rsc/g/benchgraffiti/havlak/havlak3.go
720903 720903 Total objects (flat / cumulative)
...
     .      .  277:     for i := 0; i < size; i++ {
311296 311296  278:             nodes[i] = new(UnionFindNode)
     .      .  279:     }
     .      .  280:
     .      .  281:     // Step a:
     .      .  282:     //   - initialize all nodes as unvisited.
     .      .  283:     //   - depth-first traversal and numbering.
     .      .  284:     //   - unreached BB's are marked as dead.
     .      .  285:     //
     .      .  286:     for i, bb := range cfgraph.Blocks {
     .      .  287:             number[bb.Name] = unvisited
409600 409600  288:             nonBackPreds[i] = make(map[int]bool)
     .      .  289:     }
...
(pprof)Copy

func appendUnique(a []int, x int) []int {
    for _, y := range a {
        if x == y {
            return a
        }
    }
    return append(a, x)
}Copy

$ make havlak4
go build havlak4.go
$ ./xtime ./havlak4
# of loops: 76000 (including 1 artificial root node)
11.84u 0.08s 11.94r 810416kB ./havlak4
$Copy

$ make havlak4.prof
./havlak4 -cpuprofile=havlak4.prof
# of loops: 76000 (including 1 artificial root node)
$ go tool pprof havlak4 havlak4.prof
Welcome to pprof!  For help, type 'help'.
(pprof) top10
Total: 1173 samples
     205  17.5%  17.5%     1083  92.3% main.FindLoops
     138  11.8%  29.2%      215  18.3% scanblock
      88   7.5%  36.7%       96   8.2% sweepspan
      76   6.5%  43.2%      597  50.9% runtime.mallocgc
      75   6.4%  49.6%       78   6.6% runtime.settype_flush
      74   6.3%  55.9%       75   6.4% flushptrbuf
      64   5.5%  61.4%       64   5.5% runtime.memmove
      63   5.4%  66.8%      524  44.7% runtime.growslice
      51   4.3%  71.1%       51   4.3% main.DFS
      50   4.3%  75.4%      146  12.4% runtime.MCache_Alloc
(pprof)Copy

(pprof) web mallocgcCopy

$ go tool pprof --nodefraction=0.1 havlak4 havlak4.prof
Welcome to pprof!  For help, type 'help'.
(pprof) web mallocgcCopy

(pprof) list FindLoops
...
     .      .  270: func FindLoops(cfgraph *CFG, lsgraph *LSG) {
     .      .  271:     if cfgraph.Start == nil {
     .      .  272:             return
     .      .  273:     }
     .      .  274:
     .      .  275:     size := cfgraph.NumNodes()
     .      .  276:
     .    145  277:     nonBackPreds := make([][]int, size)
     .      9  278:     backPreds := make([][]int, size)
     .      .  279:
     .      1  280:     number := make([]int, size)
     .     17  281:     header := make([]int, size, size)
     .      .  282:     types := make([]int, size, size)
     .      .  283:     last := make([]int, size, size)
     .      .  284:     nodes := make([]*UnionFindNode, size, size)
     .      .  285:
     .      .  286:     for i := 0; i < size; i++ {
     2     79  287:             nodes[i] = new(UnionFindNode)
     .      .  288:     }
...
(pprof)Copy

var cache struct {
    size int
    nonBackPreds [][]int
    backPreds [][]int
    number []int
    header []int
    types []int
    last []int
    nodes []*UnionFindNode
}Copy

if cache.size < size {
    cache.size = size
    cache.nonBackPreds = make([][]int, size)
    cache.backPreds = make([][]int, size)
    cache.number = make([]int, size)
    cache.header = make([]int, size)
    cache.types = make([]int, size)
    cache.last = make([]int, size)
    cache.nodes = make([]*UnionFindNode, size)
    for i := range cache.nodes {
        cache.nodes[i] = new(UnionFindNode)
    }
}

nonBackPreds := cache.nonBackPreds[:size]
for i := range nonBackPreds {
    nonBackPreds[i] = nonBackPreds[i][:0]
}
backPreds := cache.backPreds[:size]
for i := range nonBackPreds {
    backPreds[i] = backPreds[i][:0]
}
number := cache.number[:size]
header := cache.header[:size]
types := cache.types[:size]
last := cache.last[:size]
nodes := cache.nodes[:size]Copy

$ make havlak5
go build havlak5.go
$ ./xtime ./havlak5
# of loops: 76000 (including 1 artificial root node)
8.03u 0.06s 8.11r 770352kB ./havlak5
$Copy

$ make havlak6
go build havlak6.go
$ ./xtime ./havlak6
# of loops: 76000 (including 1 artificial root node)
2.26u 0.02s 2.29r 360224kB ./havlak6
$Copy

$ ./xtime ./havlak6 -reuseloopgraph=false
# of loops: 76000 (including 1 artificial root node)
3.69u 0.06s 3.76r 797120kB ./havlak6 -reuseloopgraph=false
$Copy

$ make havlak6cc
g++ -O3 -o havlak6cc havlak6.cc
$ ./xtime ./havlak6cc
# of loops: 76000 (including 1 artificial root node)
1.99u 0.19s 2.19r 387936kB ./havlak6ccCopy

$ wc havlak6.cc; wc havlak6.go
 401 1220 9040 havlak6.cc
 461 1441 9467 havlak6.go
$Copy

import _ "net/http/pprof"Copy

go tool pprof http://localhost:6060/debug/pprof/profile   # 30-second CPU profile
go tool pprof http://localhost:6060/debug/pprof/heap      # heap profile
go tool pprof http://localhost:6060/debug/pprof/block     # goroutine blocking profileCopy

本文章首发在 LearnKu.com 网站上。