cmd/6c: Optimize rotate expressions to use rotate instructions.

For simplicity, only recognizes expressions of the exact form
"(x << a) | (x >> b)" where x is a variable and a and b are
integer constant expressions that add to x's bit width.

Fixes #4629.

$ cat rotate.c
unsigned int
rotate(unsigned int x)
{
        x = (x << 3) | (x >> (sizeof(x) * 8 - 3));
        return x;
}

## BEFORE
$ go tool 6c -S rotate.c
(rotate.c:2)	TEXT	rotate+0(SB),$0-8
(rotate.c:2)	MOVL	x+0(FP),!!DX
(rotate.c:4)	MOVL	DX,!!AX
(rotate.c:4)	SALL	$3,!!AX
(rotate.c:4)	MOVL	DX,!!CX
(rotate.c:4)	SHRL	$29,!!CX
(rotate.c:4)	ORL	CX,!!AX
(rotate.c:5)	RET	,!!
(rotate.c:5)	RET	,!!
(rotate.c:5)	END	,!!

## AFTER
$ go tool 6c -S rotate.c
(rotate.c:2)	TEXT	rotate+0(SB),$0-8
(rotate.c:4)	MOVL	x+0(FP),!!AX
(rotate.c:4)	ROLL	$3,!!AX
(rotate.c:5)	RET	,!!
(rotate.c:5)	RET	,!!
(rotate.c:5)	END	,!!

R=rsc, minux.ma
CC=golang-dev
https://golang.org/cl/7069056

src/cmd/6c/cgen.c

@@ -265,6 +265,18 @@ cgen(Node *n, Node *nn)
 				break;
 			}
 		}
+		if(n->op == OOR && l->op == OASHL && r->op == OLSHR
+		&& l->right->op == OCONST && r->right->op == OCONST
+		&& l->left->op == ONAME && r->left->op == ONAME
+		&& l->left->sym == r->left->sym
+		&& l->right->vconst + r->right->vconst == 8 * l->left->type->width) {
+			regalloc(&nod, l->left, nn);
+			cgen(l->left, &nod);
+			gopcode(OROTL, n->type, l->right, &nod);
+			gmove(&nod, nn);
+			regfree(&nod);
+			break;
+		}
 		if(n->op == OADD && l->op == OASHL && l->right->op == OCONST
 		&& (r->op != OCONST || r->vconst < -128 || r->vconst > 127)) {
 			c = l->right->vconst;

src/cmd/6c/txt.c

@@ -1360,6 +1360,16 @@ gopcode(int o, Type *ty, Node *f, Node *t)
 			a = ASALQ;
 		break;
 
+	case OROTL:
+		a = AROLL;
+		if(et == TCHAR || et == TUCHAR)
+			a = AROLB;
+		if(et == TSHORT || et == TUSHORT)
+			a = AROLW;
+		if(et == TVLONG || et == TUVLONG || et == TIND)
+			a = AROLQ;
+		break;
+
 	case OFUNC:
 		a = ACALL;
 		break;

src/cmd/8c/cgen.c

@@ -277,6 +277,18 @@ cgen(Node *n, Node *nn)
 				break;
 			}
 		}
+		if(n->op == OOR && l->op == OASHL && r->op == OLSHR
+		&& l->right->op == OCONST && r->right->op == OCONST
+		&& l->left->op == ONAME && r->left->op == ONAME
+		&& l->left->sym == r->left->sym
+		&& l->right->vconst + r->right->vconst == 8 * l->left->type->width) {
+			regalloc(&nod, l->left, nn);
+			cgen(l->left, &nod);
+			gopcode(OROTL, n->type, l->right, &nod);
+			gmove(&nod, nn);
+			regfree(&nod);
+			break;
+		}
 		if(n->op == OADD && l->op == OASHL && l->right->op == OCONST
 		&& (r->op != OCONST || r->vconst < -128 || r->vconst > 127)) {
 			c = l->right->vconst;

src/cmd/8c/txt.c

@@ -1253,6 +1253,14 @@ gopcode(int o, Type *ty, Node *f, Node *t)
 			a = ASALW;
 		break;
 
+	case OROTL:
+		a = AROLL;
+		if(et == TCHAR || et == TUCHAR)
+			a = AROLB;
+		if(et == TSHORT || et == TUSHORT)
+			a = AROLW;
+		break;
+
 	case OFUNC:
 		a = ACALL;
 		break;

src/cmd/cc/cc.h

@@ -325,6 +325,7 @@ enum
 	OINDEX,
 	OFAS,
 	OREGPAIR,
+	OROTL,
 
 	OEND
 };

src/cmd/cc/sub.c

@@ -1515,6 +1515,7 @@ Init	onamesinit[] =
 	OINDEX,		0,	"INDEX",
 	OFAS,		0,	"FAS",
 	OREGPAIR,	0,	"REGPAIR",
+	OROTL,		0,	"ROTL",
 	OEND,		0,	"END",
 	-1,		0,	0,
 };