freebsd-dev/lib/libc/arm/string/memcmp.S
Olivier Houchard 3e3aced335 Fix the align-to-32-bits code.
Obtained from:	NetBSD
2004-09-23 23:10:56 +00:00

181 lines
6.1 KiB
ArmAsm

/* $NetBSD: memcmp.S,v 1.3 2003/10/14 07:51:45 scw Exp $ */
/*
* Copyright 2003 Wasabi Systems, Inc.
* All rights reserved.
*
* Written by Steve C. Woodford for Wasabi Systems, Inc.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed for the NetBSD Project by
* Wasabi Systems, Inc.
* 4. The name of Wasabi Systems, Inc. may not be used to endorse
* or promote products derived from this software without specific prior
* written permission.
*
* THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WASABI SYSTEMS, INC
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/
/*
* Copyright (c) 2002 ARM Ltd
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. The name of the company may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY ARM LTD ``AS IS'' AND ANY EXPRESS OR IMPLIED
* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL ARM LTD BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <machine/asm.h>
__FBSDID("$FreeBSD$");
ENTRY(memcmp)
mov ip, r0
#if defined(_KERNEL) && !defined(_STANDALONE)
cmp r2, #0x06
beq .Lmemcmp_6bytes
#endif
mov r0, #0x00
/* Are both addresses aligned the same way? */
cmp r2, #0x00
eornes r3, ip, r1
moveq pc, lr /* len == 0, or same addresses! */
tst r3, #0x03
subne r2, r2, #0x01
bne .Lmemcmp_bytewise2 /* Badly aligned. Do it the slow way */
/* Word-align the addresses, if necessary */
sub r3, r1, #0x05
ands r3, r3, #0x03
add r3, r3, r3, lsl #1
addne pc, pc, r3, lsl #3
nop
/* Compare up to 3 bytes */
ldrb r0, [ip], #0x01
ldrb r3, [r1], #0x01
subs r0, r0, r3
movne pc, lr
subs r2, r2, #0x01
moveq pc, lr
/* Compare up to 2 bytes */
ldrb r0, [ip], #0x01
ldrb r3, [r1], #0x01
subs r0, r0, r3
movne pc, lr
subs r2, r2, #0x01
moveq pc, lr
/* Compare 1 byte */
ldrb r0, [ip], #0x01
ldrb r3, [r1], #0x01
subs r0, r0, r3
movne pc, lr
subs r2, r2, #0x01
moveq pc, lr
/* Compare 4 bytes at a time, if possible */
subs r2, r2, #0x04
bcc .Lmemcmp_bytewise
.Lmemcmp_word_aligned:
ldr r0, [ip], #0x04
ldr r3, [r1], #0x04
subs r2, r2, #0x04
cmpcs r0, r3
beq .Lmemcmp_word_aligned
sub r0, r0, r3
/* Correct for extra subtraction, and check if done */
adds r2, r2, #0x04
cmpeq r0, #0x00 /* If done, did all bytes match? */
moveq pc, lr /* Yup. Just return */
/* Re-do the final word byte-wise */
sub ip, ip, #0x04
sub r1, r1, #0x04
.Lmemcmp_bytewise:
add r2, r2, #0x03
.Lmemcmp_bytewise2:
ldrb r0, [ip], #0x01
ldrb r3, [r1], #0x01
subs r2, r2, #0x01
cmpcs r0, r3
beq .Lmemcmp_bytewise2
sub r0, r0, r3
mov pc, lr
#if defined(_KERNEL) && !defined(_STANDALONE)
/*
* 6 byte compares are very common, thanks to the network stack.
* This code is hand-scheduled to reduce the number of stalls for
* load results. Everything else being equal, this will be ~32%
* faster than a byte-wise memcmp.
*/
.align 5
.Lmemcmp_6bytes:
ldrb r3, [r1, #0x00] /* r3 = b2#0 */
ldrb r0, [ip, #0x00] /* r0 = b1#0 */
ldrb r2, [r1, #0x01] /* r2 = b2#1 */
subs r0, r0, r3 /* r0 = b1#0 - b2#0 */
ldreqb r3, [ip, #0x01] /* r3 = b1#1 */
movne pc, lr /* Return if mismatch on #0 */
subs r0, r3, r2 /* r0 = b1#1 - b2#1 */
ldreqb r3, [r1, #0x02] /* r3 = b2#2 */
ldreqb r0, [ip, #0x02] /* r0 = b1#2 */
movne pc, lr /* Return if mismatch on #1 */
ldrb r2, [r1, #0x03] /* r2 = b2#3 */
subs r0, r0, r3 /* r0 = b1#2 - b2#2 */
ldreqb r3, [ip, #0x03] /* r3 = b1#3 */
movne pc, lr /* Return if mismatch on #2 */
subs r0, r3, r2 /* r0 = b1#3 - b2#3 */
ldreqb r3, [r1, #0x04] /* r3 = b2#4 */
ldreqb r0, [ip, #0x04] /* r0 = b1#4 */
movne pc, lr /* Return if mismatch on #3 */
ldrb r2, [r1, #0x05] /* r2 = b2#5 */
subs r0, r0, r3 /* r0 = b1#4 - b2#4 */
ldreqb r3, [ip, #0x05] /* r3 = b1#5 */
movne pc, lr /* Return if mismatch on #4 */
sub r0, r3, r2 /* r0 = b1#5 - b2#5 */
mov pc, lr
#endif