Merge pull request #13 from Jille/memset-generic-faster

Speed up non-asm Memset implementation 3x
bwesterb · Jul 22, 2024 · 3b33e86 · 3b33e86
2 parents 3560730 + 1f7abc9
commit 3b33e86
Showing 1 changed file with 13 additions and 1 deletion.
diff --git a/lib.go b/lib.go
@@ -111,7 +111,19 @@ func Memset(dst []byte, b byte) {
 }
 
 func memsetGeneric(dst []byte, b byte) {
-	for i := range dst {
+	if b == 0 {
+		// Special case that the Go compiler can optimize.
+		for i := range dst {
+			dst[i] = 0
+		}
+		return
+	}
+	eightB := 0x0101010101010101 * uint64(b)
+	i := 0
+	for ; i <= len(dst)-8; i += 8 {
+		binary.LittleEndian.PutUint64(dst[i:], eightB)
+	}
+	for ; i < len(dst); i++ {
 		dst[i] = b
 	}
 }