diff --git a/src/unicode/utf8/utf8.go b/src/unicode/utf8/utf8.go
index 6938c7e6a7..1e9f666e23 100644
--- a/src/unicode/utf8/utf8.go
+++ b/src/unicode/utf8/utf8.go
@@ -475,6 +475,11 @@ func RuneStart(b byte) bool { return b&0xC0 != 0x80 }
 
 // Valid reports whether p consists entirely of valid UTF-8-encoded runes.
 func Valid(p []byte) bool {
+	// This optimization avoids the need to recompute the capacity
+	// when generating code for p[8:], bringing it to parity with
+	// ValidString, which was 20% faster on long ASCII strings.
+	p = p[:len(p):len(p)]
+
 	// Fast path. Check for and skip 8 bytes of ASCII characters per iteration.
 	for len(p) >= 8 {
 		// Combining two 32 bit loads allows the same code to be used
diff --git a/src/unicode/utf8/utf8_test.go b/src/unicode/utf8/utf8_test.go
index e9be4d2d63..e7c31222cc 100644
--- a/src/unicode/utf8/utf8_test.go
+++ b/src/unicode/utf8/utf8_test.go
@@ -6,6 +6,7 @@ package utf8_test
 
 import (
 	"bytes"
+	"strings"
 	"testing"
 	"unicode"
 	. "unicode/utf8"
@@ -554,6 +555,8 @@ func BenchmarkRuneCountInStringTenJapaneseChars(b *testing.B) {
 	}
 }
 
+var ascii100000 = strings.Repeat("0123456789", 10000)
+
 func BenchmarkValidTenASCIIChars(b *testing.B) {
 	s := []byte("0123456789")
 	for i := 0; i < b.N; i++ {
@@ -561,12 +564,32 @@ func BenchmarkValidTenASCIIChars(b *testing.B) {
 	}
 }
 
+func BenchmarkValid100KASCIIChars(b *testing.B) {
+	s := []byte(ascii100000)
+	for i := 0; i < b.N; i++ {
+		Valid(s)
+	}
+}
+
 func BenchmarkValidTenJapaneseChars(b *testing.B) {
 	s := []byte("日本語日本語日本語日")
 	for i := 0; i < b.N; i++ {
 		Valid(s)
 	}
 }
+func BenchmarkValidLongMostlyASCII(b *testing.B) {
+	longMostlyASCII := []byte(longStringMostlyASCII)
+	for i := 0; i < b.N; i++ {
+		Valid(longMostlyASCII)
+	}
+}
+
+func BenchmarkValidLongJapanese(b *testing.B) {
+	longJapanese := []byte(longStringJapanese)
+	for i := 0; i < b.N; i++ {
+		Valid(longJapanese)
+	}
+}
 
 func BenchmarkValidStringTenASCIIChars(b *testing.B) {
 	for i := 0; i < b.N; i++ {
@@ -574,12 +597,47 @@ func BenchmarkValidStringTenASCIIChars(b *testing.B) {
 	}
 }
 
+func BenchmarkValidString100KASCIIChars(b *testing.B) {
+	for i := 0; i < b.N; i++ {
+		ValidString(ascii100000)
+	}
+}
+
 func BenchmarkValidStringTenJapaneseChars(b *testing.B) {
 	for i := 0; i < b.N; i++ {
 		ValidString("日本語日本語日本語日")
 	}
 }
 
+func BenchmarkValidStringLongMostlyASCII(b *testing.B) {
+	for i := 0; i < b.N; i++ {
+		ValidString(longStringMostlyASCII)
+	}
+}
+
+func BenchmarkValidStringLongJapanese(b *testing.B) {
+	for i := 0; i < b.N; i++ {
+		ValidString(longStringJapanese)
+	}
+}
+
+var longStringMostlyASCII string // ~100KB, ~97% ASCII
+var longStringJapanese string    // ~100KB, non-ASCII
+
+func init() {
+	const japanese = "日本語日本語日本語日"
+	var b bytes.Buffer
+	for i := 0; b.Len() < 100_000; i++ {
+		if i%100 == 0 {
+			b.WriteString(japanese)
+		} else {
+			b.WriteString("0123456789")
+		}
+	}
+	longStringMostlyASCII = b.String()
+	longStringJapanese = strings.Repeat(japanese, 100_000/len(japanese))
+}
+
 func BenchmarkEncodeASCIIRune(b *testing.B) {
 	buf := make([]byte, UTFMax)
 	for i := 0; i < b.N; i++ {