関連記事
GitHub - devlights/blog-summary: ブログ「いろいろ備忘録日記」のまとめ
概要
以下、自分用のメモです。
Goではよくチャネルを利用しますが、送受信(ch<-, <-ch)自体がどれくらいの速度で動くのかというのがちょっと気になったのでベンチマーク取りました。流れるデータに対しての処理は何もしないのもちょっとと思ったので、シンプルな加算処理だけしてます。
あくまで自分用のメモですが、忘れたときに見れるようにここにメモしておきます。
サンプル
benchmark_test.go
package main import ( "fmt" "io" "sync" "testing" ) func BenchmarkChanSync(b *testing.B) { benchcases := []struct { Name string In int }{ {"100", 100}, {"1000", 1000}, {"10000", 10000}, {"100000", 100000}, {"1000000", 1000000}, } for _, bc := range benchcases { bc := bc b.Run(bc.Name, func(b *testing.B) { for b.Loop() { var ( ch = make(chan int, bc.In) ) for i := range bc.In { ch <- i } close(ch) var ( total uint64 ) for v := range ch { total += uint64(v) } fmt.Fprintln(io.Discard, total) } }) } } func BenchmarkChanAsync(b *testing.B) { benchcases := []struct { Name string In int }{ {"100", 100}, {"1000", 1000}, {"10000", 10000}, {"100000", 100000}, {"1000000", 1000000}, } for _, bc := range benchcases { bc := bc b.Run(bc.Name, func(b *testing.B) { for b.Loop() { var ( ch = make(chan int, bc.In) wg = sync.WaitGroup{} ) wg.Add(2) go func(ch chan<- int) { defer wg.Done() defer close(ch) for i := range bc.In { ch <- i } }(ch) go func(ch <-chan int) { defer wg.Done() var ( total uint64 ) for v := range ch { total += uint64(v) } fmt.Fprintln(io.Discard, total) }(ch) wg.Wait() } }) } }
Taskfile.yml
# https://taskfile.dev version: '3' tasks: default: cmds: - task: clean - task: run - task: stat run: cmds: - go test -bench . -benchmem -count 10 | tee bench.out stat: cmds: - benchstat bench.out clean: cmds: - rm -f bench.out status: - test ! -f bench.out
shell
実行すると以下のように出力されます。
$ task run task: [run] go test -bench . -count 10 | tee bench.out goos: linux goarch: amd64 pkg: github.com/devlights/try-golang/examples/benchmarks/channel_send_recv cpu: AMD EPYC 7B13 BenchmarkChanSync/100-16 279804 4031 ns/op 1032 B/op 2 allocs/op BenchmarkChanSync/100-16 302364 4064 ns/op 1032 B/op 2 allocs/op BenchmarkChanSync/100-16 274219 4072 ns/op 1032 B/op 2 allocs/op BenchmarkChanSync/100-16 290083 4071 ns/op 1032 B/op 2 allocs/op BenchmarkChanSync/100-16 294525 4030 ns/op 1032 B/op 2 allocs/op BenchmarkChanSync/100-16 287929 4288 ns/op 1032 B/op 2 allocs/op BenchmarkChanSync/100-16 276304 4064 ns/op 1032 B/op 2 allocs/op BenchmarkChanSync/100-16 292827 4047 ns/op 1032 B/op 2 allocs/op BenchmarkChanSync/100-16 295461 3967 ns/op 1032 B/op 2 allocs/op BenchmarkChanSync/100-16 291420 4059 ns/op 1032 B/op 2 allocs/op BenchmarkChanSync/1000-16 30254 39316 ns/op 8205 B/op 2 allocs/op BenchmarkChanSync/1000-16 29754 39928 ns/op 8205 B/op 2 allocs/op BenchmarkChanSync/1000-16 30517 39586 ns/op 8205 B/op 2 allocs/op BenchmarkChanSync/1000-16 30960 39365 ns/op 8205 B/op 2 allocs/op BenchmarkChanSync/1000-16 27843 41077 ns/op 8205 B/op 2 allocs/op BenchmarkChanSync/1000-16 30912 38975 ns/op 8205 B/op 2 allocs/op BenchmarkChanSync/1000-16 30784 38748 ns/op 8205 B/op 2 allocs/op BenchmarkChanSync/1000-16 30766 38676 ns/op 8205 B/op 2 allocs/op BenchmarkChanSync/1000-16 29863 39719 ns/op 8205 B/op 2 allocs/op BenchmarkChanSync/1000-16 31249 38666 ns/op 8205 B/op 2 allocs/op BenchmarkChanSync/10000-16 2983 389806 ns/op 81983 B/op 2 allocs/op BenchmarkChanSync/10000-16 3055 384617 ns/op 81980 B/op 2 allocs/op BenchmarkChanSync/10000-16 3229 391278 ns/op 81981 B/op 2 allocs/op BenchmarkChanSync/10000-16 3098 387064 ns/op 81981 B/op 2 allocs/op BenchmarkChanSync/10000-16 3178 383327 ns/op 81981 B/op 2 allocs/op BenchmarkChanSync/10000-16 3002 386414 ns/op 81981 B/op 2 allocs/op BenchmarkChanSync/10000-16 3109 383222 ns/op 81982 B/op 2 allocs/op BenchmarkChanSync/10000-16 2918 389171 ns/op 81981 B/op 2 allocs/op BenchmarkChanSync/10000-16 3013 387681 ns/op 81981 B/op 2 allocs/op BenchmarkChanSync/10000-16 3048 383782 ns/op 81981 B/op 2 allocs/op BenchmarkChanSync/100000-16 316 3972062 ns/op 803420 B/op 2 allocs/op BenchmarkChanSync/100000-16 286 3940878 ns/op 803423 B/op 2 allocs/op BenchmarkChanSync/100000-16 312 3853476 ns/op 803411 B/op 2 allocs/op BenchmarkChanSync/100000-16 312 3890827 ns/op 803413 B/op 2 allocs/op BenchmarkChanSync/100000-16 315 3846943 ns/op 803424 B/op 2 allocs/op BenchmarkChanSync/100000-16 306 3919488 ns/op 803423 B/op 2 allocs/op BenchmarkChanSync/100000-16 307 3916240 ns/op 803419 B/op 2 allocs/op BenchmarkChanSync/100000-16 294 3941979 ns/op 803423 B/op 2 allocs/op BenchmarkChanSync/100000-16 309 3969874 ns/op 803425 B/op 2 allocs/op BenchmarkChanSync/100000-16 304 3870358 ns/op 803413 B/op 2 allocs/op BenchmarkChanSync/1000000-16 28 38731244 ns/op 8004792 B/op 3 allocs/op BenchmarkChanSync/1000000-16 31 38448564 ns/op 8004835 B/op 3 allocs/op BenchmarkChanSync/1000000-16 28 38591415 ns/op 8004808 B/op 3 allocs/op BenchmarkChanSync/1000000-16 31 38557124 ns/op 8004810 B/op 3 allocs/op BenchmarkChanSync/1000000-16 27 39037175 ns/op 8004824 B/op 3 allocs/op BenchmarkChanSync/1000000-16 30 39187820 ns/op 8004822 B/op 3 allocs/op BenchmarkChanSync/1000000-16 30 38997505 ns/op 8004781 B/op 3 allocs/op BenchmarkChanSync/1000000-16 31 38464896 ns/op 8004857 B/op 3 allocs/op BenchmarkChanSync/1000000-16 30 39552623 ns/op 8004766 B/op 3 allocs/op BenchmarkChanSync/1000000-16 26 38956326 ns/op 8004833 B/op 3 allocs/op BenchmarkChanAsync/100-16 186831 5874 ns/op 1161 B/op 7 allocs/op BenchmarkChanAsync/100-16 201355 5977 ns/op 1161 B/op 7 allocs/op BenchmarkChanAsync/100-16 203440 6075 ns/op 1160 B/op 7 allocs/op BenchmarkChanAsync/100-16 205980 6014 ns/op 1160 B/op 7 allocs/op BenchmarkChanAsync/100-16 204482 6194 ns/op 1160 B/op 7 allocs/op BenchmarkChanAsync/100-16 190962 5961 ns/op 1160 B/op 7 allocs/op BenchmarkChanAsync/100-16 197955 6332 ns/op 1160 B/op 7 allocs/op BenchmarkChanAsync/100-16 180174 6139 ns/op 1160 B/op 7 allocs/op BenchmarkChanAsync/100-16 202866 6156 ns/op 1160 B/op 7 allocs/op BenchmarkChanAsync/100-16 202098 6012 ns/op 1160 B/op 7 allocs/op BenchmarkChanAsync/1000-16 26412 45228 ns/op 8334 B/op 7 allocs/op BenchmarkChanAsync/1000-16 24825 48751 ns/op 8334 B/op 7 allocs/op BenchmarkChanAsync/1000-16 27230 44634 ns/op 8334 B/op 7 allocs/op BenchmarkChanAsync/1000-16 26161 45000 ns/op 8334 B/op 7 allocs/op BenchmarkChanAsync/1000-16 27258 45609 ns/op 8334 B/op 7 allocs/op BenchmarkChanAsync/1000-16 24712 46048 ns/op 8334 B/op 7 allocs/op BenchmarkChanAsync/1000-16 26526 45107 ns/op 8334 B/op 7 allocs/op BenchmarkChanAsync/1000-16 26499 45045 ns/op 8335 B/op 7 allocs/op BenchmarkChanAsync/1000-16 24843 47015 ns/op 8335 B/op 7 allocs/op BenchmarkChanAsync/1000-16 26824 44760 ns/op 8334 B/op 7 allocs/op BenchmarkChanAsync/10000-16 2098 536619 ns/op 82120 B/op 7 allocs/op BenchmarkChanAsync/10000-16 2265 532042 ns/op 82120 B/op 7 allocs/op BenchmarkChanAsync/10000-16 2313 534333 ns/op 82119 B/op 7 allocs/op BenchmarkChanAsync/10000-16 2124 540292 ns/op 82118 B/op 7 allocs/op BenchmarkChanAsync/10000-16 2268 527942 ns/op 82119 B/op 7 allocs/op BenchmarkChanAsync/10000-16 2200 539408 ns/op 82119 B/op 7 allocs/op BenchmarkChanAsync/10000-16 2232 536863 ns/op 82120 B/op 7 allocs/op BenchmarkChanAsync/10000-16 2238 539517 ns/op 82119 B/op 7 allocs/op BenchmarkChanAsync/10000-16 2214 528823 ns/op 82119 B/op 7 allocs/op BenchmarkChanAsync/10000-16 2320 528874 ns/op 82120 B/op 7 allocs/op BenchmarkChanAsync/100000-16 200 5951932 ns/op 803600 B/op 8 allocs/op BenchmarkChanAsync/100000-16 194 6225372 ns/op 803622 B/op 8 allocs/op BenchmarkChanAsync/100000-16 200 5940656 ns/op 803594 B/op 8 allocs/op BenchmarkChanAsync/100000-16 199 6100676 ns/op 803604 B/op 8 allocs/op BenchmarkChanAsync/100000-16 198 6042795 ns/op 803597 B/op 8 allocs/op BenchmarkChanAsync/100000-16 196 6151063 ns/op 803593 B/op 8 allocs/op BenchmarkChanAsync/100000-16 194 6200071 ns/op 803609 B/op 8 allocs/op BenchmarkChanAsync/100000-16 196 6236949 ns/op 803610 B/op 8 allocs/op BenchmarkChanAsync/100000-16 195 6140092 ns/op 803647 B/op 8 allocs/op BenchmarkChanAsync/100000-16 190 6236760 ns/op 803595 B/op 8 allocs/op BenchmarkChanAsync/1000000-16 19 60881378 ns/op 8005094 B/op 9 allocs/op BenchmarkChanAsync/1000000-16 19 60285705 ns/op 8005045 B/op 9 allocs/op BenchmarkChanAsync/1000000-16 19 58995496 ns/op 8005061 B/op 9 allocs/op BenchmarkChanAsync/1000000-16 19 61019145 ns/op 8005164 B/op 10 allocs/op BenchmarkChanAsync/1000000-16 18 63189106 ns/op 8005026 B/op 9 allocs/op BenchmarkChanAsync/1000000-16 20 60307970 ns/op 8005048 B/op 9 allocs/op BenchmarkChanAsync/1000000-16 18 61124251 ns/op 8005091 B/op 9 allocs/op BenchmarkChanAsync/1000000-16 18 61482151 ns/op 8005060 B/op 9 allocs/op BenchmarkChanAsync/1000000-16 20 60296411 ns/op 8005076 B/op 9 allocs/op BenchmarkChanAsync/1000000-16 20 60038558 ns/op 8005068 B/op 9 allocs/op PASS ok github.com/devlights/try-golang/examples/benchmarks/channel_send_recv 118.281s
benchstat で見ると以下のような感じ。
$ task stat task: [stat] benchstat bench.out goos: linux goarch: amd64 pkg: github.com/devlights/try-golang/examples/benchmarks/channel_send_recv cpu: AMD EPYC 7B13 │ bench.out │ │ sec/op │ ChanSync/100-16 4.062µ ± 1% ChanSync/1000-16 39.34µ ± 2% ChanSync/10000-16 386.7µ ± 1% ChanSync/100000-16 3.918m ± 2% ChanSync/1000000-16 38.84m ± 1% ChanAsync/100-16 6.045µ ± 2% ChanAsync/1000-16 45.17µ ± 4% ChanAsync/10000-16 535.5µ ± 1% ChanAsync/100000-16 6.146m ± 3% ChanAsync/1000000-16 60.59m ± 1% geomean 468.7µ │ bench.out │ │ B/op │ ChanSync/100-16 1.008Ki ± 0% ChanSync/1000-16 8.013Ki ± 0% ChanSync/10000-16 80.06Ki ± 0% ChanSync/100000-16 784.6Ki ± 0% ChanSync/1000000-16 7.634Mi ± 0% ChanAsync/100-16 1.133Ki ± 0% ChanAsync/1000-16 8.139Ki ± 0% ChanAsync/10000-16 80.19Ki ± 0% ChanAsync/100000-16 784.8Ki ± 0% ChanAsync/1000000-16 7.634Mi ± 0% geomean 84.24Ki │ bench.out │ │ allocs/op │ ChanSync/100-16 2.000 ± 0% ChanSync/1000-16 2.000 ± 0% ChanSync/10000-16 2.000 ± 0% ChanSync/100000-16 2.000 ± 0% ChanSync/1000000-16 3.000 ± 0% ChanAsync/100-16 7.000 ± 0% ChanAsync/1000-16 7.000 ± 0% ChanAsync/10000-16 7.000 ± 0% ChanAsync/100000-16 8.000 ± 0% ChanAsync/1000000-16 9.000 ± 0% geomean 4.049
結果をどう捉えるのかは個々人の感覚になりますが、個人的には同期・非同期での利用に関わらず充分なパフォーマンスです。
参考情報
Goのおすすめ書籍
過去の記事については、以下のページからご参照下さい。
サンプルコードは、以下の場所で公開しています。
