Golang slice append vs assign performance

Question

To make slice append operation faster we need to allocate enough capacity. There\'s two ways to append slice, Here is the code:

func BenchmarkSliceAppend(b *t         


        

Answer 1

It seems like some improvements of Go compiler or runtime have been introduced since this question has been posted, so now (Go 1.10.1) there is no significant difference between append and direct assignment by index.

Also, I had to change your benchmarks slightly because of OOM panics.

package main

import "testing"

var result []int

const size = 32

const iterations = 100 * 1000 * 1000

func doAssign() {
    data := make([]int, size)
    for i := 0; i < size; i++ {
        data[i] = i
    }
    result = data
}

func doAppend() {
    data := make([]int, 0, size)
    for i := 0; i < size; i++ {
        data = append(data, i)
    }
    result = data
}

func BenchmarkAssign(b *testing.B) {
    b.N = iterations
    for i := 0; i < b.N; i++ {
        doAssign()
    }
}

func BenchmarkAppend(b *testing.B) {
    b.N = iterations
    for i := 0; i < b.N; i++ {
        doAppend()
    }
}

Results:

➜  bench_slice_assign go test -bench=Bench .
goos: linux
goarch: amd64
BenchmarkAssign-4       100000000           80.9 ns/op
BenchmarkAppend-4       100000000           81.9 ns/op
PASS
ok      _/home/isaev/troubles/bench_slice_assign    16.288s