Improve benchmark, "do no harm" optimization.

Author: ncruces

src/cmp/bench_test.go

@@ -2,43 +2,56 @@ package cmp_test
 
 import (
 	"cmp"
-	"slices"
-	"strconv"
+	"math"
 	"testing"
 )
 
+var sum int
+
 func BenchmarkCompare_int(b *testing.B) {
-	var lst [1_000_000]int
-	for i := range lst {
-		lst[i] = i
+	lst := [...]int{
+		0xfa3, 0x7fe, 0x03c, 0xcb9,
+		0x4ce, 0x4fb, 0x7d5, 0x38f,
+		0x73b, 0x322, 0x85c, 0xf4d,
+		0xbbc, 0x032, 0x059, 0xb93,
 	}
-	b.ResetTimer()
-
 	for n := 0; n < b.N; n++ {
-		slices.SortFunc(lst[:], cmp.Compare)
+		sum += cmp.Compare(lst[n%len(lst)], lst[(2*n)%len(lst)])
 	}
 }
 
 func BenchmarkCompare_float64(b *testing.B) {
-	var lst [1_000_000]float64
-	for i := range lst {
-		lst[i] = float64(i)
+	lst := [...]float64{
+		0.35573281, 0.77552566, 0.19006500, 0.66436280,
+		0.02769279, 0.97572397, 0.40945068, 0.26422857,
+		0.10985792, 0.35659522, 0.82752613, 0.18875522,
+		0.16410543, 0.03578153, 0.51636871, math.NaN(),
 	}
-	b.ResetTimer()
-
 	for n := 0; n < b.N; n++ {
-		slices.SortFunc(lst[:], cmp.Compare)
+		sum += cmp.Compare(lst[n%len(lst)], lst[(2*n)%len(lst)])
 	}
 }
 
-func BenchmarkCompare_string(b *testing.B) {
-	var lst [1_000_000]string
-	for i := range lst {
-		lst[i] = strconv.Itoa(i)
+func BenchmarkCompare_strings(b *testing.B) {
+	lst := [...]string{
+		"time",
+		"person",
+		"year",
+		"way",
+		"day",
+		"thing",
+		"man",
+		"world",
+		"life",
+		"hand",
+		"part",
+		"child",
+		"eye",
+		"woman",
+		"place",
+		"work",
 	}
-	b.ResetTimer()
-
 	for n := 0; n < b.N; n++ {
-		slices.SortFunc(lst[:], cmp.Compare)
+		sum += cmp.Compare(lst[n%len(lst)], lst[(2*n)%len(lst)])
 	}
 }

src/cmp/cmp.go

@@ -38,19 +38,19 @@ func Less[T Ordered](x, y T) bool {
 // For floating-point types, a NaN is considered less than any non-NaN,
 // a NaN is considered equal to a NaN, and -0.0 is equal to 0.0.
 func Compare[T Ordered](x, y T) int {
-	if x == y {
-		return 0
-	}
-	if x < y {
-		return -1
-	}
-	if isNaN(x) {
-		if isNaN(y) {
-			return 0
+	if x != y {
+		if isNaN(x) {
+			if isNaN(y) {
+				return 0
+			}
+			return -1
+		}
+		if x < y {
+			return -1
 		}
-		return -1
+		return +1
 	}
-	return +1
+	return 0
 }
 
 // isNaN reports whether x is a NaN without requiring the math package.