• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1/*
2 * Copyright (c) 2017 ARM Ltd
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 3. The name of the company may not be used to endorse or promote
14 *    products derived from this software without specific prior written
15 *    permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY ARM LTD ``AS IS'' AND ANY EXPRESS OR IMPLIED
18 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
19 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL ARM LTD BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
21 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
22 * TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
23 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
24 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
25 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
26 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29/* Assumptions:
30 *
31 * ARMv8-a, AArch64, unaligned accesses.
32 */
33
34#include <private/bionic_asm.h>
35
36#define L(l) .L ## l
37
38/* Parameters and result.  */
39#define src1		x0
40#define src2		x1
41#define limit		x2
42#define result		w0
43
44/* Internal variables.  */
45#define data1		x3
46#define data1w		w3
47#define data1h		x4
48#define data2		x5
49#define data2w		w5
50#define data2h		x6
51#define tmp1		x7
52#define tmp2		x8
53
54/* Small inputs of less than 8 bytes are handled separately.  This allows the
55   main code to be speed up using unaligned loads since there are now at least
56   8 bytes to be compared.  If the first 8 bytes are equal, align src1.
57   This ensures each iteration does at most one unaligned access even if both
58   src1 and src2 are unaligned, and mutually aligned inputs behave as if
59   aligned.  After the main loop, process the last 16 bytes using unaligned
60   accesses.  */
61
62ENTRY(memcmp)
63.p2align 6
64	subs	limit, limit, 8
65	b.lo	L(less8)
66
67	/* Limit >= 8, so check first 8 bytes using unaligned loads.  */
68	ldr	data1, [src1], 8
69	ldr	data2, [src2], 8
70	cmp	data1, data2
71	b.ne	L(return)
72
73	subs	limit, limit, 8
74	b.gt	L(more16)
75
76	ldr	data1, [src1, limit]
77	ldr	data2, [src2, limit]
78	b	L(return)
79
80L(more16):
81	ldr	data1, [src1], 8
82	ldr	data2, [src2], 8
83	cmp	data1, data2
84	bne	L(return)
85
86	/* Jump directly to comparing the last 16 bytes for 32 byte (or less)
87	   strings.  */
88	subs	limit, limit, 16
89	b.ls	L(last_bytes)
90
91	/* We overlap loads between 0-32 bytes at either side of SRC1 when we
92	   try to align, so limit it only to strings larger than 128 bytes.  */
93	cmp	limit, 96
94	b.ls	L(loop16)
95
96	/* Align src1 and adjust src2 with bytes not yet done.  */
97	and	tmp1, src1, 15
98	add	limit, limit, tmp1
99	sub	src1, src1, tmp1
100	sub	src2, src2, tmp1
101
102	/* Loop performing 16 bytes per iteration using aligned src1.
103	   Limit is pre-decremented by 16 and must be larger than zero.
104	   Exit if <= 16 bytes left to do or if the data is not equal.  */
105	.p2align 4
106L(loop16):
107	ldp	data1, data1h, [src1], 16
108	ldp	data2, data2h, [src2], 16
109	subs	limit, limit, 16
110	ccmp	data1, data2, 0, hi
111	ccmp	data1h, data2h, 0, eq
112	b.eq	L(loop16)
113
114	cmp	data1, data2
115	bne	L(return)
116	mov	data1, data1h
117	mov	data2, data2h
118	cmp	data1, data2
119	bne	L(return)
120
121	/* Compare last 1-16 bytes using unaligned access.  */
122L(last_bytes):
123	add	src1, src1, limit
124	add	src2, src2, limit
125	ldp	data1, data1h, [src1]
126	ldp	data2, data2h, [src2]
127	cmp     data1, data2
128	bne	L(return)
129	mov	data1, data1h
130	mov	data2, data2h
131	cmp	data1, data2
132
133	/* Compare data bytes and set return value to 0, -1 or 1.  */
134L(return):
135#ifndef __AARCH64EB__
136	rev	data1, data1
137	rev	data2, data2
138#endif
139	cmp     data1, data2
140L(ret_eq):
141	cset	result, ne
142	cneg	result, result, lo
143	ret
144
145	.p2align 4
146	/* Compare up to 8 bytes.  Limit is [-8..-1].  */
147L(less8):
148	adds	limit, limit, 4
149	b.lo	L(less4)
150	ldr	data1w, [src1], 4
151	ldr	data2w, [src2], 4
152	cmp	data1w, data2w
153	b.ne	L(return)
154	sub	limit, limit, 4
155L(less4):
156	adds	limit, limit, 4
157	beq	L(ret_eq)
158L(byte_loop):
159	ldrb	data1w, [src1], 1
160	ldrb	data2w, [src2], 1
161	subs	limit, limit, 1
162	ccmp	data1w, data2w, 0, ne	/* NZCV = 0b0000.  */
163	b.eq	L(byte_loop)
164	sub	result, data1w, data2w
165	ret
166
167END(memcmp)
168