1// Derived from Inferno's libkern/memmove-386.s (adapted for amd64)
2// https://bitbucket.org/inferno-os/inferno-os/src/default/libkern/memmove-386.s
3//
4//         Copyright © 1994-1999 Lucent Technologies Inc. All rights reserved.
5//         Revisions Copyright © 2000-2007 Vita Nuova Holdings Limited (www.vitanuova.com).  All rights reserved.
6//         Portions Copyright 2009 The Go Authors. All rights reserved.
7//
8// Permission is hereby granted, free of charge, to any person obtaining a copy
9// of this software and associated documentation files (the "Software"), to deal
10// in the Software without restriction, including without limitation the rights
11// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
12// copies of the Software, and to permit persons to whom the Software is
13// furnished to do so, subject to the following conditions:
14//
15// The above copyright notice and this permission notice shall be included in
16// all copies or substantial portions of the Software.
17//
18// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
19// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
20// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
21// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
22// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
23// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24// THE SOFTWARE.
25
26#include "textflag.h"
27
28// void runtime·memmove(void*, void*, uintptr)
29TEXT runtime·memmove(SB), NOSPLIT, $0-24
30
31	MOVQ	to+0(FP), DI
32	MOVQ	from+8(FP), SI
33	MOVQ	n+16(FP), BX
34
35	// REP instructions have a high startup cost, so we handle small sizes
36	// with some straightline code. The REP MOVSQ instruction is really fast
37	// for large sizes. The cutover is approximately 1K.
38tail:
39	TESTQ	BX, BX
40	JEQ	move_0
41	CMPQ	BX, $2
42	JBE	move_1or2
43	CMPQ	BX, $4
44	JBE	move_3or4
45	CMPQ	BX, $8
46	JB	move_5through7
47	JE	move_8
48	CMPQ	BX, $16
49	JBE	move_9through16
50
51/*
52 * check and set for backwards
53 */
54	CMPQ	SI, DI
55	JLS	back
56
57/*
58 * forward copy loop
59 */
60forward:
61	MOVQ	BX, CX
62	SHRQ	$3, CX
63	ANDQ	$7, BX
64
65	REP;	MOVSQ
66	JMP	tail
67
68back:
69/*
70 * check overlap
71 */
72	MOVQ	SI, CX
73	ADDQ	BX, CX
74	CMPQ	CX, DI
75	JLS	forward
76
77/*
78 * whole thing backwards has
79 * adjusted addresses
80 */
81	ADDQ	BX, DI
82	ADDQ	BX, SI
83	STD
84
85/*
86 * copy
87 */
88	MOVQ	BX, CX
89	SHRQ	$3, CX
90	ANDQ	$7, BX
91
92	SUBQ	$8, DI
93	SUBQ	$8, SI
94	REP;	MOVSQ
95
96	CLD
97	ADDQ	$8, DI
98	ADDQ	$8, SI
99	SUBQ	BX, DI
100	SUBQ	BX, SI
101	JMP	tail
102
103move_1or2:
104	MOVB	(SI), AX
105	MOVB	-1(SI)(BX*1), CX
106	MOVB	AX, (DI)
107	MOVB	CX, -1(DI)(BX*1)
108	RET
109move_0:
110	RET
111move_3or4:
112	MOVW	(SI), AX
113	MOVW	-2(SI)(BX*1), CX
114	MOVW	AX, (DI)
115	MOVW	CX, -2(DI)(BX*1)
116	RET
117move_5through7:
118	MOVL	(SI), AX
119	MOVL	-4(SI)(BX*1), CX
120	MOVL	AX, (DI)
121	MOVL	CX, -4(DI)(BX*1)
122	RET
123move_8:
124	// We need a separate case for 8 to make sure we write pointers atomically.
125	MOVQ	(SI), AX
126	MOVQ	AX, (DI)
127	RET
128move_9through16:
129	MOVQ	(SI), AX
130	MOVQ	-8(SI)(BX*1), CX
131	MOVQ	AX, (DI)
132	MOVQ	CX, -8(DI)(BX*1)
133	RET
134