pass by value vs pass by reference

alloc_test.go

package main

import "testing"

type Stuff struct {
    Hello   string
    Goodbye string
    Slice   []string
}

func NewStuff() Stuff {
    return Stuff{
        Hello:   "hello!",
        Goodbye: "goodbye!",
        Slice:   []string{"dogs", "are", "pretty", "cool"},
    }
}

func BenchmarkPassByValue(b *testing.B) {
    stuff := NewStuff()
    for i := 0; i < b.N; i++ {
        byValue(stuff)
    }
}

func BenchmarkPassByPointer(b *testing.B) {
    stuff := NewStuff()
    for i := 0; i < b.N; i++ {
        byPointer(&stuff)
    }
}

func byPointer(stuff *Stuff) {
}

func byValue(stuff Stuff) {
}

bench results

BenchmarkPassByValue    500000000                7.87 ns/op
BenchmarkPassByPointer  2000000000               0.30 ns/op

A fairer comparison in many circumstances would be:

func BenchmarkPassByValue(b *testing.B) {
for i := 0; i < b.N; i++ {
stuff := NewStuff()
byValue(stuff)
}
}

func BenchmarkPassByPointer(b *testing.B) {
for i := 0; i < b.N; i++ {
stuff := NewStuff()
byPointer(&stuff)
}
}

because passing by pointer often implies you must create
a pointer to pass. The difference then is considerably
smaller (i see 42.7ns vs 34.2).

And if the pointer escapes (as is not uncommon) then the
by-value option is considerably better (50ns vs 109ns)

I used this code:
http://play.golang.org/p/B17c5pBWH6