Ok, I think there are two reasons why the copy version is slower:
1. the copy version initializes each element twice actutally.
2. alloc a new slice by using make is slow than by using append. But why
this?
package main
import (
"testing"
)
const N = 1024 * 1024 // number of elements
type Element int64
var x = make([]Element, N)
var y []Element
func Benchmark_AllocWithMake(b *testing.B) {
for i := 0; i < b.N; i++ {
y = make([]Element, N)
}
}
func Benchmark_AllocWithAppend(b *testing.B) {
for i := 0; i < b.N; i++ {
y = append([]Element(nil), x...)
}
}
func Benchmark_CloneWithCopy(b *testing.B) {
for i := 0; i < b.N; i++ {
t := make([]Element, N)
copy(t, x)
y = t
}
}
func Benchmark_CloneWithAppend(b *testing.B) {
for i := 0; i < b.N; i++ {
y = append([]Element(nil), x...)
}
}
Benchmark_AllocWithMake-4 1000 1223666 ns/op
Benchmark_AllocWithAppend-4 2000 725361 ns/op
Benchmark_CloneWithCopy-4 1000 1813321 ns/op
Benchmark_CloneWithAppend-4 2000 735011 ns/op
On Tuesday, June 27, 2017 at 10:58:28 PM UTC+8, T L wrote:
>
>
> package main
>
> import (
> "testing"
> )
>
> const N = 1024 * 1024 // number of elements
> type Element int64
> var x = make([]Element, N)
> var y []Element
>
>
> func Benchmark_CloneWithAppend(b *testing.B) {
> for i := 0; i < b.N; i++ {
> y = append([]Element(nil), x...)
> }
> }
>
> func Benchmark_CloneWithCopy(b *testing.B) {
> b.ResetTimer()
> for i := 0; i < b.N; i++ {
> t := make([]Element, N)
> copy(t, x)
> y = t
> }
> }
>
> /*
> $ go test -bench=.
> Benchmark_CloneWithAppend-4 2000 738281 ns/op
> Benchmark_CloneWithCopy-4 1000 1822462 ns/op
> */
>
--
You received this message because you are subscribed to the Google Groups
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
