Random Number Generation

rand.IntN(100) from math/rand/v2 (Go 1.22+). The global functions use a ChaCha8-based generator with cheap per-thread state instead of a global lock, and the range reduction algorithm avoids v1's modulo bias trick. Faster, safer seeding by default, and the idiomatic choice for all non-cryptographic randomness.

// randRange is the exclusive upper bound for the generated random numbers.
const randRange = 100

func BenchmarkMathRandV2_run(b *testing.B) {
	for i := 0; i < b.N; i++ {
		sink = rand.IntN(randRange)
	}
}

rand.Intn(100) from the legacy math/rand. The global functions share one generator behind a mutex, so concurrent use contends, and the underlying algorithm is a decades-old additive generator. Kept around for compatibility; new code should not use it.

// randRange is the exclusive upper bound for the generated random numbers.
const randRange = 100

func BenchmarkMathRand_run(b *testing.B) {
	for i := 0; i < b.N; i++ {
		sink = rand.Intn(randRange)
	}
}

rand.Int(rand.Reader, max) from crypto/rand, which reads from the operating system's CSPRNG and allocates big.Int values along the way. Orders of magnitude slower — and completely worth it whenever the random value is security-relevant: session tokens, keys, nonces, password salts.

// randRange is the exclusive upper bound for the generated random numbers.
const randRange = 100

func BenchmarkCryptoRand_run(b *testing.B) {
	upperBound := big.NewInt(randRange)
	b.ResetTimer()
	for i := 0; i < b.N; i++ {
		n, _ := rand.Int(rand.Reader, upperBound)
		sink = int(n.Int64())
	}
}