runtime: minor string/rune optimizations

Eliminate a spill in concatstrings.
Provide bounds elim hints in runetochar.
No significant benchmark movement.

Before:
"".runetochar t=1 size=412 args=0x28 locals=0x0
"".concatstrings t=1 size=736 args=0x30 locals=0x98

After:
"".runetochar t=1 size=337 args=0x28 locals=0x0
"".concatstrings t=1 size=711 args=0x30 locals=0x90

Change-Id: Icce646976cb20a223163b7e72a54761193ac17e3
Reviewed-on: https://go-review.googlesource.com/27460
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Martin Möhrmann <martisch@uos.de>
TryBot-Result: Gobot Gobot <gobot@golang.org>

src/runtime/rune.go

@@ -178,6 +178,7 @@ func runetochar(str []byte, r rune) int {
 	 *	0080-07FF => t2 tx
 	 */
 	if c <= rune2 {
+		_ = str[1]
 		str[0] = byte(t2 | (c >> (1 * bitx)))
 		str[1] = byte(tx | (c & maskx))
 		return 2
@@ -201,6 +202,7 @@ func runetochar(str []byte, r rune) int {
 	 *	0800-FFFF => t3 tx tx
 	 */
 	if c <= rune3 {
+		_ = str[2]
 		str[0] = byte(t3 | (c >> (2 * bitx)))
 		str[1] = byte(tx | ((c >> (1 * bitx)) & maskx))
 		str[2] = byte(tx | (c & maskx))
@@ -211,6 +213,7 @@ func runetochar(str []byte, r rune) int {
 	 * four character sequence (21-bit value)
 	 *     10000-1FFFFF => t4 tx tx tx
 	 */
+	_ = str[3]
 	str[0] = byte(t4 | (c >> (3 * bitx)))
 	str[1] = byte(tx | ((c >> (2 * bitx)) & maskx))
 	str[2] = byte(tx | ((c >> (1 * bitx)) & maskx))

src/runtime/string.go

@@ -47,10 +47,9 @@ func concatstrings(buf *tmpBuf, a []string) string {
 		return a[idx]
 	}
 	s, b := rawstringtmp(buf, l)
-	l = 0
 	for _, x := range a {
-		copy(b[l:], x)
-		l += len(x)
+		copy(b, x)
+		b = b[len(x):]
 	}
 	return s
 }