I benchmarked golang simple countup of not fixed code, mutex, atomic.AddInt64 or atomic.Value on goroutine.
Result
fastest result: atomic.Value > mutex > not fixed code > atomic.AddInt64
One time result
$ go test -bench . goos: linux goarch: amd64 BenchmarkInc_Countup-4 2000000 957 ns/op --- BENCH: BenchmarkInc_Countup-4 acounter_test.go:13: 1 -> 1 acounter_test.go:13: 100 -> 100 acounter_test.go:13: 10000 -> 9752 acounter_test.go:13: 1000000 -> 982973 acounter_test.go:13: 2000000 -> 1962072 BenchmarkLock_Countup-4 2000000 936 ns/op --- BENCH: BenchmarkLock_Countup-4 acounter_test.go:20: 1 -> 1 acounter_test.go:20: 100 -> 100 acounter_test.go:20: 10000 -> 10000 acounter_test.go:20: 1000000 -> 1000000 acounter_test.go:20: 2000000 -> 2000000 BenchmarkAtomic_Countup-4 2000000 943 ns/op --- BENCH: BenchmarkAtomic_Countup-4 acounter_test.go:27: 1 -> 1 acounter_test.go:27: 100 -> 100 acounter_test.go:27: 10000 -> 10000 acounter_test.go:27: 1000000 -> 1000000 acounter_test.go:27: 2000000 -> 2000000 BenchmarkAtomicValue_Countup-4 2000000 904 ns/op --- BENCH: BenchmarkAtomicValue_Countup-4 acounter_test.go:36: 1 -> 1 acounter_test.go:36: 100 -> 54 acounter_test.go:36: 10000 -> 7125 acounter_test.go:36: 1000000 -> 791697 acounter_test.go:36: 2000000 -> 1586774 PASS Three times result
$ for i in $(seq 1 3); do go test -bench . | grep 'ns/op'; done BenchmarkInc_Countup-4 2000000 998 ns/op BenchmarkLock_Countup-4 2000000 986 ns/op BenchmarkAtomic_Countup-4 2000000 1130 ns/op BenchmarkAtomicValue_Countup-4 2000000 987 ns/op BenchmarkInc_Countup-4 2000000 989 ns/op BenchmarkLock_Countup-4 1000000 1113 ns/op BenchmarkAtomic_Countup-4 1000000 1010 ns/op BenchmarkAtomicValue_Countup-4 2000000 1013 ns/op BenchmarkInc_Countup-4 1000000 1018 ns/op BenchmarkLock_Countup-4 1000000 1311 ns/op BenchmarkAtomic_Countup-4 1000000 1051 ns/op BenchmarkAtomicValue_Countup-4 1000000 1125 ns/op 
Top comments (1)
Hi YooWaan! You can replace bash command to go-test arguments
instead of using this:
you can use this: