|
|
1.1 root 1: /*
2: * Copyright (c) 1999 Apple Computer, Inc. All rights reserved.
3: *
4: * @APPLE_LICENSE_HEADER_START@
5: *
6: * Portions Copyright (c) 1999 Apple Computer, Inc. All Rights
7: * Reserved. This file contains Original Code and/or Modifications of
8: * Original Code as defined in and that are subject to the Apple Public
9: * Source License Version 1.1 (the "License"). You may not use this file
10: * except in compliance with the License. Please obtain a copy of the
11: * License at http://www.apple.com/publicsource and read it before using
12: * this file.
13: *
14: * The Original Code and all software distributed under the License are
15: * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
16: * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
17: * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
18: * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the
19: * License for the specific language governing rights and limitations
20: * under the License.
21: *
22: * @APPLE_LICENSE_HEADER_END@
23: */
24:
25: /* Copyright (c) 1991 NeXT Computer, Inc. All rights reserved.
26: *
27: * File: libc/i386/ansi/memmove.c
28: * Author: Bruce Martin, NeXT Computer, Inc.
29: *
30: * This file contains machine dependent code for block copies
31: * on NeXT i386-based products. Currently tuned for the i486.
32: *
33: * HISTORY
34: * 9-Aug-92 Bruce Martin ([email protected])
35: * Created.
36: */
37:
38:
39: /*
40: * 486 Factoids:
41: * - if (move <= 16 bytes) use unrolled mem-mem mov instructions
42: * - if (move > 16 bytes) use "rep movs" [this avoids prefetch stalls]
43: * - Aligned read is better than aligned write. Both aligned is even better.
44: */
45:
46:
47:
48: #define MIN_TO_ALIGN 16
49:
50: /*
51: * Convert a void * into a unsigned int so arithmetic can be done
52: * (Technically, the gnu C compiler is tolerant of arithmetic
53: * on void *, but ansi isn't; so we do this.)
54: */
55: #define UNS(voidp) ((unsigned int)(voidp))
56:
57: /*
58: * Number of bytes addr is past an integer alignment
59: */
60: #define BYTES_PAST_INT_ALIGNMENT(voidp) (UNS(voidp) & (sizeof(int) - 1))
61:
62: #if defined(NX_CURRENT_COMPILER_RELEASE) && (NX_CURRENT_COMPILER_RELEASE < 320)
63: #define SIREG "e"
64: #else
65: #define SIREG "S"
66: #endif
67:
68:
69: static inline void
70: simple_fwd_char_copy(void *dst, const void *src, int len)
71: {
72: asm("rep; movsb"
73: : /* no outputs */
74: : "&c" (len), "D" (dst), SIREG (src)
75: : "ecx", "esi", "edi");
76: }
77:
78: static inline void
79: simple_bkwd_char_copy(void *dst, const void *src, int len)
80: {
81: /* NOTE: assumes DF flag is set */
82: UNS(src)--;
83: UNS(dst)--;
84: asm("rep; movsb"
85: : /* no outputs */
86: : "&c" (len), "D" (dst), SIREG (src)
87: : "ecx", "esi", "edi");
88: }
89:
90:
91: static inline void
92: simple_fwd_int_copy(void *dst, const void *src, int len)
93: {
94: asm("rep; movsl"
95: : /* no outputs */
96: : "&c" (len>>2), "D" (dst), SIREG (src)
97: : "ecx", "esi", "edi");
98: }
99:
100: static inline void
101: simple_bkwd_int_copy(void *dst, const void *src, int len)
102: {
103: /* NOTE: assumes DF flag is set */
104: UNS(dst) -= sizeof(int);
105: UNS(src) -= sizeof(int);
106: asm("rep; movsl"
107: : /* no outputs */
108: : "&c" (len>>2), "D" (dst), SIREG (src)
109: : "ecx", "esi", "edi");
110: }
111:
112:
113: void *memmove(void *, const void *, unsigned long);
114:
115: void bcopy(const void *src, void *dst, unsigned long ulen)
116: {
117: (void) memmove(dst, src, ulen);
118: }
119:
120: void ovbcopy(const void *src, void *dst, unsigned long ulen)
121: {
122: (void) memmove(dst, src, ulen);
123: }
124:
125:
126: void *memmove(void *dst, const void *src, unsigned long ulen)
127: {
128: int len = ulen;
129: int need_to_move;
130: void *orig_dst = dst;
131:
132: if ((dst <= src) || (src + ulen <= dst)) {
133: /* Forward copy */
134:
135: if (len < MIN_TO_ALIGN) {
136: simple_fwd_char_copy(dst, src, len);
137: return orig_dst;
138: }
139:
140: if (need_to_move = BYTES_PAST_INT_ALIGNMENT(src)) {
141: need_to_move = sizeof(int) - need_to_move;
142: simple_fwd_char_copy(dst, src, need_to_move);
143: len -= need_to_move;
144: UNS(dst) += need_to_move;
145: UNS(src) += need_to_move;
146: }
147:
148: simple_fwd_int_copy(dst, src, len);
149:
150: if (need_to_move = (len & 3)) {
151: UNS(src) += (len & ~3);
152: UNS(dst) += (len & ~3);
153: simple_fwd_char_copy(dst, src, need_to_move);
154: }
155: } else {
156: /* Backward copy */
157:
158: UNS(src) += len;
159: UNS(dst) += len;
160:
161: /* Set decrement bit */
162: asm("std");
163:
164: if (len <= MIN_TO_ALIGN) {
165: simple_bkwd_char_copy(dst, src, len);
166: /* Clear decrement bit */
167: asm("cld");
168: return orig_dst;
169: }
170:
171: if (need_to_move = BYTES_PAST_INT_ALIGNMENT(src)) {
172: simple_bkwd_char_copy(dst, src, need_to_move);
173: len -= need_to_move;
174: UNS(dst) -= need_to_move;
175: UNS(src) -= need_to_move;
176: }
177:
178: simple_bkwd_int_copy(dst, src, len);
179:
180: if (need_to_move = (len & 3)) {
181: UNS(src) -= (len & ~3);
182: UNS(dst) -= (len & ~3);
183: simple_bkwd_char_copy(dst, src, need_to_move);
184: }
185:
186: /* Clear decrement bit */
187: asm("cld");
188: }
189: return orig_dst;
190: }
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.