Why does the runtime of my program change when I use go run vs go build and executing my program?
I'm using the time package to measure the time elapsed during program execution, and I stumbled upon this behaviour.
package main
import (
"fmt"
"io/ioutil"
"time"
)
func sumInt(b []byte, c chan int) {
sum := 0
for _, num := range b {
sum += int(num)
}
c <- sum
}
func main() {
start := time.Now()
dat, err := ioutil.ReadFile("nums.txt")
if err != nil {
return
}
partSum1 := make(chan int)
partSum2 := make(chan int)
go sumInt(dat[:len(dat)/2], partSum1)
go sumInt(dat[len(dat)/2:], partSum2)
sum := <-partSum1 + <-partSum2
fmt.Println(sum)
elapsed := time.Since(start)
fmt.Printf("Elapsed %s
", elapsed)
}
❱ go run runtime.go
2502263
Elapsed 133.893µs
❱ go build runtime.go
❱ ./runtime
2502263
Elapsed 402.435µs
❱
In Go, we don't favor meaningless microbenchmarks. If you are going to run a benchmark, use the Go testing package.
For example,
Output:
$ go test sum_test.go -bench=. -benchmem
goos: linux
goarch: amd64
BenchmarkSum-8 500000 2905 ns/op 4288 B/op 3 allocs/op
$
sum_test.go:
package main
import (
"testing"
)
func sumInt(b []byte, c chan int) {
sum := 0
for _, num := range b {
sum += int(num)
}
c <- sum
}
func BenchmarkSum(b *testing.B) {
for N := 0; N < b.N; N++ {
dat := make([]byte, 4*1024)
partSum1 := make(chan int)
partSum2 := make(chan int)
go sumInt(dat[:len(dat)/2], partSum1)
go sumInt(dat[len(dat)/2:], partSum2)
sum := <-partSum1 + <-partSum2
_ = sum
}
}