fastrand: optimize Uint32N()

name                old time/op  new time/op  delta
FastRandBuf-16      32.5ns ± 4%  18.8ns ± 7%  -42.02%  (p=0.000 n=9+10)
FastRandUint32N-16  0.91ns ± 1%  0.19ns ±12%  -78.83%  (p=0.000 n=9+10)

ref #33239

util/fastrand/random.go

@@ -37,10 +37,9 @@ func Uint32() uint32
 
 // Uint32N returns, as an uint32, a pseudo-random number in [0,n).
 func Uint32N(n uint32) uint32 {
-	if n&(n-1) == 0 { // n is power of two, can mask
-		return Uint32() & (n - 1)
-	}
-	return Uint32() % n
+	// This is similar to Uint32() % n, but faster.
+	// See https://lemire.me/blog/2016/06/27/a-fast-alternative-to-the-modulo-reduction/
+	return uint32(uint64(Uint32()) * uint64(n) >> 32)
 }
 
 // Uint64N returns, as an uint64, a pseudo-random number in [0,n).

util/fastrand/random_test.go

@@ -42,6 +42,22 @@ func TestRand(t *testing.T) {
 	require.Less(t, sum, 24)
 }
 
+func BenchmarkFastRandBuf(b *testing.B) {
+	b.RunParallel(func(pb *testing.PB) {
+		for pb.Next() {
+			Buf(20)
+		}
+	})
+}
+
+func BenchmarkFastRandUint32N(b *testing.B) {
+	b.RunParallel(func(pb *testing.PB) {
+		for pb.Next() {
+			Uint32N(127)
+		}
+	})
+}
+
 func BenchmarkFastRand(b *testing.B) {
 	b.RunParallel(func(pb *testing.PB) {
 		for pb.Next() {