Go Benchmark function

main.go

…

Expected output

Lesson notes

Benchmark function

Learn the func BenchmarkXxx signature and go test -bench.

Benchmark functions

Benchmarks live in a _test.go file, start with Benchmark and take a *testing.B. The measured code runs b.N times; Go raises b.N automatically until the measurement is reliable.

func BenchmarkBirlestir(b *testing.B) {
    for i := 0; i < b.N; i++ {
        strings.Join([]string{"a", "b", "c"}, ",")
    }
}

Run: go test -bench=. -benchmem → BenchmarkBirlestir-8 25000000 48 ns/op 16 B/op 1 allocs/op

Not fooling the compiler

A computation whose result is never used can be removed by the compiler, so the benchmark measures nothing. Assign the result to a package-level variable:

var sink string

func BenchmarkX(b *testing.B) {
    for i := 0; i < b.N; i++ {
        sink = strings.Repeat("a", 10)
    }
}

Since Go 1.24 the for b.Loop() { ... } form solves this problem and the timer handling by itself.

Exercise First benchmark

Run a benchmark with testing.Benchmark that calls fmt.Sprintf("%d", i); assign the result to a sink variable. Print whether the result's N is greater than 0.

Show example solution

package main

import (
    "fmt"
    "testing"
)

var sink string

func main() {
    r := testing.Benchmark(func(b *testing.B) {
        for i := 0; i < b.N; i++ {
            sink = fmt.Sprintf("%d", i)
        }
    })
    fmt.Println(r.N > 0)
}

Test yourself

What does the <code>-8</code> suffix in the output show?
- The iteration count
- The allocation count
- The Go version
- The GOMAXPROCS value
Show answer

The GOMAXPROCS value — The number at the end of the benchmark name says how many Ps the measurement used; -cpu=1,4 tries different values.

Iteration count

for i := 0; i < b.____; i++ {
    sink = hesapla()
}

Show answer

for i := 0; i < b.N; i++ {
    sink = hesapla()
}

Output column

ns/op

Time per operation
B/op

Bytes per operation
allocs/op

Allocations per operation
-8

GOMAXPROCS

Next lesson: b.ResetTimer