runtime: replace runtime.mcpy with runtime.memmove

faster string operations, and more

tested on linux/386

runtime_test.BenchmarkSliceToString                    642          532  -17.13%
runtime_test.BenchmarkStringToSlice                    636          528  -16.98%
runtime_test.BenchmarkConcatString                    1109          897  -19.12%

R=r, iant, rsc
CC=golang-dev
https://golang.org/cl/4674042

src/cmd/cgo/out.go

@@ -708,7 +708,7 @@ void
 ·_Cfunc_CString(String s, int8 *p)
 {
 	p = runtime·cmalloc(s.len+1);
-	runtime·mcpy((byte*)p, s.str, s.len);
+	runtime·memmove((byte*)p, s.str, s.len);
 	p[s.len] = 0;
 	FLUSH(&p);
 }

src/pkg/runtime/386/closure.c

@@ -45,7 +45,7 @@ runtime·closure(int32 siz, byte *fn, byte *arg0)
 	q = p + n - siz;
 
 	if(siz > 0) {
-		runtime·mcpy(q, (byte*)&arg0, siz);
+		runtime·memmove(q, (byte*)&arg0, siz);
 
 		// SUBL $siz, SP
 		*p++ = 0x81;

src/pkg/runtime/386/memmove.s

@@ -27,9 +27,6 @@ TEXT runtime·memmove(SB), 7, $0
 	MOVL	to+0(FP), DI
 	MOVL	fr+4(FP), SI
 	MOVL	n+8(FP), BX
-	CMPL	BX, $0
-	JLT	fault
-
 /*
  * check and set for backwards
  */
@@ -87,12 +84,3 @@ back:
 	MOVL	to+0(FP),AX
 	RET
 
-/*
- * if called with negative count,
- * treat as error rather than
- * rotating all of memory
- */
-fault:
-	MOVL	$0,SI
-	MOVL	0(SI), AX
-	RET

src/pkg/runtime/amd64/closure.c

@@ -45,7 +45,7 @@ runtime·closure(int32 siz, byte *fn, byte *arg0)
 	q = p + n - siz;
 
 	if(siz > 0) {
-		runtime·mcpy(q, (byte*)&arg0, siz);
+		runtime·memmove(q, (byte*)&arg0, siz);
 
 		// SUBQ $siz, SP
 		*p++ = 0x48;

src/pkg/runtime/amd64/memmove.s

@@ -28,8 +28,6 @@ TEXT runtime·memmove(SB), 7, $0
 	MOVQ	to+0(FP), DI
 	MOVQ	fr+8(FP), SI
 	MOVLQSX	n+16(FP), BX
-	CMPQ	BX, $0
-	JLT	fault
 
 /*
  * check and set for backwards
@@ -88,12 +86,3 @@ back:
 	MOVQ	to+0(FP),AX
 	RET
 
-/*
- * if called with negative count,
- * treat as error rather than
- * rotating all of memory
- */
-fault:
-	MOVQ	$0,SI
-	MOVQ	0(SI), AX
-	RET

src/pkg/runtime/arm/closure.c

@@ -83,7 +83,7 @@ runtime·closure(int32 siz, byte *fn, byte *arg0)
 	*pc++ = 0xe52de000 | (siz + 4);
 
 	if(siz > 0) {
-		runtime·mcpy(q, (byte*)&arg0, siz);
+		runtime·memmove(q, (byte*)&arg0, siz);
 
 		//	MOVW	$vars(PC), R0
 		*pc = 0xe28f0000 | (int32)(q - (byte*)pc - 8);

src/pkg/runtime/hashmap.h

@@ -65,7 +65,7 @@
 
 #define	malloc		runtime·mal
 #define	memset(a,b,c)	runtime·memclr((byte*)(a), (uint32)(c))
-#define	memcpy(a,b,c)	runtime·mcpy((byte*)(a),(byte*)(b),(uint32)(c))
+#define	memcpy(a,b,c)	runtime·memmove((byte*)(a),(byte*)(b),(uint32)(c))
 #define	assert(a)	if(!(a)) runtime·throw("assert")
 #define free(x)	runtime·free(x)
 #define memmove(a,b,c)	runtime·memmove(a, b, c)

src/pkg/runtime/plan9/thread.c

@@ -47,11 +47,11 @@ runtime·exit(int32)
 		pid = pid/10;
 	}
 	p = buf;
-	runtime·mcpy((void*)p, (void*)"/proc/", 6);
+	runtime·memmove((void*)p, (void*)"/proc/", 6);
 	p += 6;
 	for(q--; q >= tmp;)
 		*p++ = *q--;
-	runtime·mcpy((void*)p, (void*)"/notepg", 7);
+	runtime·memmove((void*)p, (void*)"/notepg", 7);
 	
 	/* post interrupt note */
 	fd = runtime·open(buf, OWRITE);

src/pkg/runtime/proc.c

@@ -720,7 +720,7 @@ runtime·oldstack(void)
 	argsize = old.argsize;
 	if(argsize > 0) {
 		sp -= argsize;
-		runtime·mcpy(top->argp, sp, argsize);
+		runtime·memmove(top->argp, sp, argsize);
 	}
 	goid = old.gobuf.g->goid;	// fault if g is bad, before gogo
 	USED(goid);
@@ -802,7 +802,7 @@ runtime·newstack(void)
 	sp = (byte*)top;
 	if(argsize > 0) {
 		sp -= argsize;
-		runtime·mcpy(sp, m->moreargp, argsize);
+		runtime·memmove(sp, m->moreargp, argsize);
 	}
 	if(thechar == '5') {
 		// caller would have saved its LR below args.
@@ -903,7 +903,7 @@ runtime·newproc1(byte *fn, byte *argp, int32 narg, int32 nret, void *callerpc)
 
 	sp = newg->stackbase;
 	sp -= siz;
-	runtime·mcpy(sp, argp, narg);
+	runtime·memmove(sp, argp, narg);
 	if(thechar == '5') {
 		// caller's LR
 		sp -= sizeof(void*);
@@ -941,7 +941,7 @@ runtime·deferproc(int32 siz, byte* fn, ...)
 		d->argp = (byte*)(&fn+2);  // skip caller's saved link register
 	else
 		d->argp = (byte*)(&fn+1);
-	runtime·mcpy(d->args, d->argp, d->siz);
+	runtime·memmove(d->args, d->argp, d->siz);
 
 	d->link = g->defer;
 	g->defer = d;
@@ -968,7 +968,7 @@ runtime·deferreturn(uintptr arg0)
 	argp = (byte*)&arg0;
 	if(d->argp != argp)
 		return;
-	runtime·mcpy(argp, d->args, d->siz);
+	runtime·memmove(argp, d->args, d->siz);
 	g->defer = d->link;
 	fn = d->fn;
 	runtime·free(d);
@@ -1367,11 +1367,11 @@ os·setenv_c(String k, String v)
 		return;
 
 	arg[0] = runtime·malloc(k.len + 1);
-	runtime·mcpy(arg[0], k.str, k.len);
+	runtime·memmove(arg[0], k.str, k.len);
 	arg[0][k.len] = 0;
 
 	arg[1] = runtime·malloc(v.len + 1);
-	runtime·mcpy(arg[1], v.str, v.len);
+	runtime·memmove(arg[1], v.str, v.len);
 	arg[1][v.len] = 0;
 
 	runtime·asmcgocall(libcgo_setenv, arg);

src/pkg/runtime/runtime.c

@@ -116,17 +116,6 @@ runtime·panicstring(int8 *s)
 	runtime·panic(err);
 }
 
-void
-runtime·mcpy(byte *t, byte *f, uint32 n)
-{
-	while(n > 0) {
-		*t = *f;
-		t++;
-		f++;
-		n--;
-	}
-}
-
 int32
 runtime·mcmp(byte *s1, byte *s2, uint32 n)
 {

src/pkg/runtime/runtime.h

@@ -406,7 +406,6 @@ uint32	runtime·rnd(uint32, uint32);
 void	runtime·prints(int8*);
 void	runtime·printf(int8*, ...);
 byte*	runtime·mchr(byte*, byte, byte*);
-void	runtime·mcpy(byte*, byte*, uint32);
 int32	runtime·mcmp(byte*, byte*, uint32);
 void	runtime·memmove(void*, void*, uint32);
 void*	runtime·mal(uintptr);

src/pkg/runtime/string.goc

@@ -60,7 +60,7 @@ runtime·gostring(byte *str)
 
 	l = runtime·findnull(str);
 	s = runtime·gostringsize(l);
-	runtime·mcpy(s.str, str, l);
+	runtime·memmove(s.str, str, l);
 	return s;
 }
 
@@ -70,7 +70,7 @@ runtime·gostringn(byte *str, int32 l)
 	String s;
 
 	s = runtime·gostringsize(l);
-	runtime·mcpy(s.str, str, l);
+	runtime·memmove(s.str, str, l);
 	return s;
 }
 
@@ -113,8 +113,8 @@ runtime·catstring(String s1, String s2)
 		return s1;
 
 	s3 = runtime·gostringsize(s1.len + s2.len);
-	runtime·mcpy(s3.str, s1.str, s1.len);
+	runtime·memmove(s3.str, s1.str, s1.len);
-	runtime·mcpy(s3.str+s1.len, s2.str, s2.len);
+	runtime·memmove(s3.str+s1.len, s2.str, s2.len);
 	return s3;
 }
 
@@ -134,7 +134,7 @@ concatstring(int32 n, String *s)
 	out = runtime·gostringsize(l);
 	l = 0;
 	for(i=0; i<n; i++) {
-		runtime·mcpy(out.str+l, s[i].str, s[i].len);
+		runtime·memmove(out.str+l, s[i].str, s[i].len);
 		l += s[i].len;
 	}
 	return out;
@@ -225,14 +225,14 @@ func intstring(v int64) (s String) {
 
 func slicebytetostring(b Slice) (s String) {
 	s = runtime·gostringsize(b.len);
-	runtime·mcpy(s.str, b.array, s.len);
+	runtime·memmove(s.str, b.array, s.len);
 }
 
 func stringtoslicebyte(s String) (b Slice) {
 	b.array = runtime·mallocgc(s.len, FlagNoPointers, 1, 1);
 	b.len = s.len;
 	b.cap = s.len;
-	runtime·mcpy(b.array, s.str, s.len);
+	runtime·memmove(b.array, s.str, s.len);
 }
 
 func sliceinttostring(b Slice) (s String) {

src/pkg/runtime/symtab.c

@@ -159,7 +159,7 @@ makepath(byte *buf, int32 nbuf, byte *path)
 			break;
 		if(p > buf && p[-1] != '/')
 			*p++ = '/';
-		runtime·mcpy(p, q, len+1);
+		runtime·memmove(p, q, len+1);
 		p += len;
 	}
 }