Xscale-optimised mem* functions, contributed by Wasabi Systems.
(Note: memcmp/memset improvements also benefit non-Xscale).
memcmp() - Compare 32-bits at a time if possible. Special-case 6-byte
comparisons, for the benefit of the network stack.
memset() - More loop unrolling, plus use of 'strd' instruction,
results in > 100% speedup on Xscale.
memcpy() - Big-endian support, unrolled loops, 'strd/ldrd/pld', plus
special-cases for very common length/alignment combinations
(at least in the kernel). Benchmarks show ~50% improvment on
Xscale.
memmove() - Big-endian support. Use fast memcpy(), above, if the regions
don't overlap. Otherwise unchanged.
This commit is contained in:
scw
parent
9d9b366ce0
commit
5e7e19ec12
|
|
@ -1,5 +1,39 @@
|
|||
/* $NetBSD: memcmp.S,v 1.2 2003/04/05 23:27:15 bjh21 Exp $ */
|
||||
/* $NetBSD: memcmp.S,v 1.3 2003/10/13 19:59:24 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.
|
||||
|
|
@ -30,21 +64,117 @@
|
|||
|
||||
#include <machine/asm.h>
|
||||
|
||||
RCSID("$NetBSD: memcmp.S,v 1.2 2003/04/05 23:27:15 bjh21 Exp $")
|
||||
RCSID("$NetBSD: memcmp.S,v 1.3 2003/10/13 19:59:24 scw Exp $")
|
||||
|
||||
ENTRY(memcmp)
|
||||
/* if ((len - 1) < 0) return 0 */
|
||||
subs r2, r2, #1
|
||||
movmi r0, #0
|
||||
movmi pc, lr
|
||||
mov ip, r0
|
||||
#if defined(_KERNEL) && !defined(_STANDALONE)
|
||||
cmp r2, #0x06
|
||||
beq .Lmemcmp_6bytes
|
||||
#endif
|
||||
mov r0, #0x00
|
||||
|
||||
/* ip == last src address to compare */
|
||||
add ip, r0, r2
|
||||
1:
|
||||
ldrb r2, [r0], #1
|
||||
ldrb r3, [r1], #1
|
||||
cmp ip, r0
|
||||
cmpcs r2, r3
|
||||
beq 1b
|
||||
sub r0, r2, r3
|
||||
/* 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 */
|
||||
and r3, r1, #0x03
|
||||
rsbs 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
|
||||
|
|
|
|||
|
|
@ -1,585 +1,7 @@
|
|||
/* $NetBSD: memcpy.S,v 1.5 2003/10/09 08:54:54 ichiro Exp $ */
|
||||
/* $NetBSD: memcpy.S,v 1.6 2003/10/13 19:59:24 scw Exp $ */
|
||||
|
||||
/*-
|
||||
* Copyright (c) 1997 The NetBSD Foundation, Inc.
|
||||
* All rights reserved.
|
||||
*
|
||||
* This code is derived from software contributed to The NetBSD Foundation
|
||||
* by Neil A. Carson and Mark Brinicombe
|
||||
*
|
||||
* 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 by the NetBSD
|
||||
* Foundation, Inc. and its contributors.
|
||||
* 4. Neither the name of The NetBSD Foundation nor the names of its
|
||||
* contributors may be used to endorse or promote products derived
|
||||
* from this software without specific prior written permission.
|
||||
*
|
||||
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
|
||||
* ``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 THE FOUNDATION OR CONTRIBUTORS
|
||||
* 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>
|
||||
|
||||
/*
|
||||
* This is one fun bit of code ...
|
||||
* Some easy listening music is suggested while trying to understand this
|
||||
* code e.g. Iron Maiden
|
||||
*
|
||||
* For anyone attempting to understand it :
|
||||
*
|
||||
* The core code is implemented here with simple stubs for memcpy()
|
||||
* memmove() and bcopy().
|
||||
*
|
||||
* All local labels are prefixed with Lmemcpy_
|
||||
* Following the prefix a label starting f is used in the forward copy code
|
||||
* while a label using b is used in the backwards copy code
|
||||
* The source and destination addresses determine whether a forward or
|
||||
* backward copy is performed.
|
||||
* Separate bits of code are used to deal with the following situations
|
||||
* for both the forward and backwards copy.
|
||||
* unaligned source address
|
||||
* unaligned destination address
|
||||
* Separate copy routines are used to produce an optimised result for each
|
||||
* of these cases.
|
||||
* The copy code will use LDM/STM instructions to copy up to 32 bytes at
|
||||
* a time where possible.
|
||||
*
|
||||
* Note: r12 (aka ip) can be trashed during the function along with
|
||||
* r0-r3 although r0-r2 have defined uses i.e. src, dest, len through out.
|
||||
* Additional registers are preserved prior to use i.e. r4, r5 & lr
|
||||
*
|
||||
* Apologies for the state of the comments ;-)
|
||||
*/
|
||||
|
||||
ENTRY(memcpy)
|
||||
ENTRY_NP(memmove)
|
||||
/* Determine copy direction */
|
||||
cmp r1, r0
|
||||
|
||||
moveq r0, #0 /* Quick abort for len=0 */
|
||||
moveq pc, lr
|
||||
|
||||
/* save leaf functions having to store this away */
|
||||
stmdb sp!, {r0, lr} /* memcpy() returns dest addr */
|
||||
|
||||
bcc .Lmemcpy_backwards
|
||||
|
||||
/* start of forwards copy */
|
||||
subs r2, r2, #4
|
||||
blt .Lmemcpy_fl4 /* less than 4 bytes */
|
||||
ands r12, r0, #3
|
||||
bne .Lmemcpy_fdestul /* oh unaligned destination addr */
|
||||
ands r12, r1, #3
|
||||
bne .Lmemcpy_fsrcul /* oh unaligned source addr */
|
||||
|
||||
.Lmemcpy_ft8:
|
||||
/* We have aligned source and destination */
|
||||
subs r2, r2, #8
|
||||
blt .Lmemcpy_fl12 /* less than 12 bytes (4 from above) */
|
||||
subs r2, r2, #0x14
|
||||
blt .Lmemcpy_fl32 /* less than 32 bytes (12 from above) */
|
||||
stmdb sp!, {r4} /* borrow r4 */
|
||||
|
||||
/* blat 32 bytes at a time */
|
||||
/* XXX for really big copies perhaps we should use more registers */
|
||||
.Lmemcpy_floop32:
|
||||
ldmia r1!, {r3, r4, r12, lr}
|
||||
stmia r0!, {r3, r4, r12, lr}
|
||||
ldmia r1!, {r3, r4, r12, lr}
|
||||
stmia r0!, {r3, r4, r12, lr}
|
||||
subs r2, r2, #0x20
|
||||
bge .Lmemcpy_floop32
|
||||
|
||||
cmn r2, #0x10
|
||||
ldmgeia r1!, {r3, r4, r12, lr} /* blat a remaining 16 bytes */
|
||||
stmgeia r0!, {r3, r4, r12, lr}
|
||||
subge r2, r2, #0x10
|
||||
ldmia sp!, {r4} /* return r4 */
|
||||
|
||||
.Lmemcpy_fl32:
|
||||
adds r2, r2, #0x14
|
||||
|
||||
/* blat 12 bytes at a time */
|
||||
.Lmemcpy_floop12:
|
||||
ldmgeia r1!, {r3, r12, lr}
|
||||
stmgeia r0!, {r3, r12, lr}
|
||||
subges r2, r2, #0x0c
|
||||
bge .Lmemcpy_floop12
|
||||
|
||||
.Lmemcpy_fl12:
|
||||
adds r2, r2, #8
|
||||
blt .Lmemcpy_fl4
|
||||
|
||||
subs r2, r2, #4
|
||||
ldrlt r3, [r1], #4
|
||||
strlt r3, [r0], #4
|
||||
ldmgeia r1!, {r3, r12}
|
||||
stmgeia r0!, {r3, r12}
|
||||
subge r2, r2, #4
|
||||
|
||||
.Lmemcpy_fl4:
|
||||
/* less than 4 bytes to go */
|
||||
adds r2, r2, #4
|
||||
#ifdef __APCS_26_
|
||||
ldmeqia sp!, {r0, pc}^ /* done */
|
||||
#if !defined(__XSCALE__) || defined(_STANDALONE)
|
||||
#include "memcpy_arm.S"
|
||||
#else
|
||||
ldmeqia sp!, {r0, pc} /* done */
|
||||
#include "memcpy_xscale.S"
|
||||
#endif
|
||||
/* copy the crud byte at a time */
|
||||
cmp r2, #2
|
||||
ldrb r3, [r1], #1
|
||||
strb r3, [r0], #1
|
||||
ldrgeb r3, [r1], #1
|
||||
strgeb r3, [r0], #1
|
||||
ldrgtb r3, [r1], #1
|
||||
strgtb r3, [r0], #1
|
||||
ldmia sp!, {r0, pc}
|
||||
|
||||
/* erg - unaligned destination */
|
||||
.Lmemcpy_fdestul:
|
||||
rsb r12, r12, #4
|
||||
cmp r12, #2
|
||||
|
||||
/* align destination with byte copies */
|
||||
ldrb r3, [r1], #1
|
||||
strb r3, [r0], #1
|
||||
ldrgeb r3, [r1], #1
|
||||
strgeb r3, [r0], #1
|
||||
ldrgtb r3, [r1], #1
|
||||
strgtb r3, [r0], #1
|
||||
subs r2, r2, r12
|
||||
blt .Lmemcpy_fl4 /* less the 4 bytes */
|
||||
|
||||
ands r12, r1, #3
|
||||
beq .Lmemcpy_ft8 /* we have an aligned source */
|
||||
|
||||
/* erg - unaligned source */
|
||||
/* This is where it gets nasty ... */
|
||||
.Lmemcpy_fsrcul:
|
||||
bic r1, r1, #3
|
||||
ldr lr, [r1], #4
|
||||
cmp r12, #2
|
||||
bgt .Lmemcpy_fsrcul3
|
||||
beq .Lmemcpy_fsrcul2
|
||||
cmp r2, #0x0c
|
||||
blt .Lmemcpy_fsrcul1loop4
|
||||
sub r2, r2, #0x0c
|
||||
stmdb sp!, {r4, r5}
|
||||
|
||||
.Lmemcpy_fsrcul1loop16:
|
||||
#ifdef __ARMEB__
|
||||
mov r3, lr, lsl #8
|
||||
ldmia r1!, {r4, r5, r12, lr}
|
||||
orr r3, r3, r4, lsr #24
|
||||
mov r4, r4, lsl #8
|
||||
orr r4, r4, r5, lsr #24
|
||||
mov r5, r5, lsl #8
|
||||
orr r5, r5, r12, lsr #24
|
||||
mov r12, r12, lsl #8
|
||||
orr r12, r12, lr, lsr #24
|
||||
#else
|
||||
mov r3, lr, lsr #8
|
||||
ldmia r1!, {r4, r5, r12, lr}
|
||||
orr r3, r3, r4, lsl #24
|
||||
mov r4, r4, lsr #8
|
||||
orr r4, r4, r5, lsl #24
|
||||
mov r5, r5, lsr #8
|
||||
orr r5, r5, r12, lsl #24
|
||||
mov r12, r12, lsr #8
|
||||
orr r12, r12, lr, lsl #24
|
||||
#endif
|
||||
stmia r0!, {r3-r5, r12}
|
||||
subs r2, r2, #0x10
|
||||
bge .Lmemcpy_fsrcul1loop16
|
||||
ldmia sp!, {r4, r5}
|
||||
adds r2, r2, #0x0c
|
||||
blt .Lmemcpy_fsrcul1l4
|
||||
|
||||
.Lmemcpy_fsrcul1loop4:
|
||||
#ifdef __ARMEB__
|
||||
mov r12, lr, lsl #8
|
||||
ldr lr, [r1], #4
|
||||
orr r12, r12, lr, lsr #24
|
||||
#else
|
||||
mov r12, lr, lsr #8
|
||||
ldr lr, [r1], #4
|
||||
orr r12, r12, lr, lsl #24
|
||||
#endif
|
||||
str r12, [r0], #4
|
||||
subs r2, r2, #4
|
||||
bge .Lmemcpy_fsrcul1loop4
|
||||
|
||||
.Lmemcpy_fsrcul1l4:
|
||||
sub r1, r1, #3
|
||||
b .Lmemcpy_fl4
|
||||
|
||||
.Lmemcpy_fsrcul2:
|
||||
cmp r2, #0x0c
|
||||
blt .Lmemcpy_fsrcul2loop4
|
||||
sub r2, r2, #0x0c
|
||||
stmdb sp!, {r4, r5}
|
||||
|
||||
.Lmemcpy_fsrcul2loop16:
|
||||
#ifdef __ARMEB__
|
||||
mov r3, lr, lsl #16
|
||||
ldmia r1!, {r4, r5, r12, lr}
|
||||
orr r3, r3, r4, lsr #16
|
||||
mov r4, r4, lsl #16
|
||||
orr r4, r4, r5, lsr #16
|
||||
mov r5, r5, lsl #16
|
||||
orr r5, r5, r12, lsr #16
|
||||
mov r12, r12, lsl #16
|
||||
orr r12, r12, lr, lsr #16
|
||||
#else
|
||||
mov r3, lr, lsr #16
|
||||
ldmia r1!, {r4, r5, r12, lr}
|
||||
orr r3, r3, r4, lsl #16
|
||||
mov r4, r4, lsr #16
|
||||
orr r4, r4, r5, lsl #16
|
||||
mov r5, r5, lsr #16
|
||||
orr r5, r5, r12, lsl #16
|
||||
mov r12, r12, lsr #16
|
||||
orr r12, r12, lr, lsl #16
|
||||
#endif
|
||||
stmia r0!, {r3-r5, r12}
|
||||
subs r2, r2, #0x10
|
||||
bge .Lmemcpy_fsrcul2loop16
|
||||
ldmia sp!, {r4, r5}
|
||||
adds r2, r2, #0x0c
|
||||
blt .Lmemcpy_fsrcul2l4
|
||||
|
||||
.Lmemcpy_fsrcul2loop4:
|
||||
#ifdef __ARMEB__
|
||||
mov r12, lr, lsl #16
|
||||
ldr lr, [r1], #4
|
||||
orr r12, r12, lr, lsr #16
|
||||
#else
|
||||
mov r12, lr, lsr #16
|
||||
ldr lr, [r1], #4
|
||||
orr r12, r12, lr, lsl #16
|
||||
#endif
|
||||
str r12, [r0], #4
|
||||
subs r2, r2, #4
|
||||
bge .Lmemcpy_fsrcul2loop4
|
||||
|
||||
.Lmemcpy_fsrcul2l4:
|
||||
sub r1, r1, #2
|
||||
b .Lmemcpy_fl4
|
||||
|
||||
.Lmemcpy_fsrcul3:
|
||||
cmp r2, #0x0c
|
||||
blt .Lmemcpy_fsrcul3loop4
|
||||
sub r2, r2, #0x0c
|
||||
stmdb sp!, {r4, r5}
|
||||
|
||||
.Lmemcpy_fsrcul3loop16:
|
||||
#ifdef __ARMEB__
|
||||
mov r3, lr, lsl #24
|
||||
ldmia r1!, {r4, r5, r12, lr}
|
||||
orr r3, r3, r4, lsr #8
|
||||
mov r4, r4, lsl #24
|
||||
orr r4, r4, r5, lsr #8
|
||||
mov r5, r5, lsl #24
|
||||
orr r5, r5, r12, lsr #8
|
||||
mov r12, r12, lsl #24
|
||||
orr r12, r12, lr, lsr #8
|
||||
#else
|
||||
mov r3, lr, lsr #24
|
||||
ldmia r1!, {r4, r5, r12, lr}
|
||||
orr r3, r3, r4, lsl #8
|
||||
mov r4, r4, lsr #24
|
||||
orr r4, r4, r5, lsl #8
|
||||
mov r5, r5, lsr #24
|
||||
orr r5, r5, r12, lsl #8
|
||||
mov r12, r12, lsr #24
|
||||
orr r12, r12, lr, lsl #8
|
||||
#endif
|
||||
stmia r0!, {r3-r5, r12}
|
||||
subs r2, r2, #0x10
|
||||
bge .Lmemcpy_fsrcul3loop16
|
||||
ldmia sp!, {r4, r5}
|
||||
adds r2, r2, #0x0c
|
||||
blt .Lmemcpy_fsrcul3l4
|
||||
|
||||
.Lmemcpy_fsrcul3loop4:
|
||||
#ifdef __ARMEB__
|
||||
mov r12, lr, lsl #24
|
||||
ldr lr, [r1], #4
|
||||
orr r12, r12, lr, lsr #8
|
||||
#else
|
||||
mov r12, lr, lsr #24
|
||||
ldr lr, [r1], #4
|
||||
orr r12, r12, lr, lsl #8
|
||||
#endif
|
||||
str r12, [r0], #4
|
||||
subs r2, r2, #4
|
||||
bge .Lmemcpy_fsrcul3loop4
|
||||
|
||||
.Lmemcpy_fsrcul3l4:
|
||||
sub r1, r1, #1
|
||||
b .Lmemcpy_fl4
|
||||
|
||||
.Lmemcpy_backwards:
|
||||
add r1, r1, r2
|
||||
add r0, r0, r2
|
||||
subs r2, r2, #4
|
||||
blt .Lmemcpy_bl4 /* less than 4 bytes */
|
||||
ands r12, r0, #3
|
||||
bne .Lmemcpy_bdestul /* oh unaligned destination addr */
|
||||
ands r12, r1, #3
|
||||
bne .Lmemcpy_bsrcul /* oh unaligned source addr */
|
||||
|
||||
.Lmemcpy_bt8:
|
||||
/* We have aligned source and destination */
|
||||
subs r2, r2, #8
|
||||
blt .Lmemcpy_bl12 /* less than 12 bytes (4 from above) */
|
||||
stmdb sp!, {r4}
|
||||
subs r2, r2, #0x14 /* less than 32 bytes (12 from above) */
|
||||
blt .Lmemcpy_bl32
|
||||
|
||||
/* blat 32 bytes at a time */
|
||||
/* XXX for really big copies perhaps we should use more registers */
|
||||
.Lmemcpy_bloop32:
|
||||
ldmdb r1!, {r3, r4, r12, lr}
|
||||
stmdb r0!, {r3, r4, r12, lr}
|
||||
ldmdb r1!, {r3, r4, r12, lr}
|
||||
stmdb r0!, {r3, r4, r12, lr}
|
||||
subs r2, r2, #0x20
|
||||
bge .Lmemcpy_bloop32
|
||||
|
||||
.Lmemcpy_bl32:
|
||||
cmn r2, #0x10
|
||||
ldmgedb r1!, {r3, r4, r12, lr} /* blat a remaining 16 bytes */
|
||||
stmgedb r0!, {r3, r4, r12, lr}
|
||||
subge r2, r2, #0x10
|
||||
adds r2, r2, #0x14
|
||||
ldmgedb r1!, {r3, r12, lr} /* blat a remaining 12 bytes */
|
||||
stmgedb r0!, {r3, r12, lr}
|
||||
subge r2, r2, #0x0c
|
||||
ldmia sp!, {r4}
|
||||
|
||||
.Lmemcpy_bl12:
|
||||
adds r2, r2, #8
|
||||
blt .Lmemcpy_bl4
|
||||
subs r2, r2, #4
|
||||
ldrlt r3, [r1, #-4]!
|
||||
strlt r3, [r0, #-4]!
|
||||
ldmgedb r1!, {r3, r12}
|
||||
stmgedb r0!, {r3, r12}
|
||||
subge r2, r2, #4
|
||||
|
||||
.Lmemcpy_bl4:
|
||||
/* less than 4 bytes to go */
|
||||
adds r2, r2, #4
|
||||
ldmeqia sp!, {r0, pc}
|
||||
|
||||
/* copy the crud byte at a time */
|
||||
cmp r2, #2
|
||||
ldrb r3, [r1, #-1]!
|
||||
strb r3, [r0, #-1]!
|
||||
ldrgeb r3, [r1, #-1]!
|
||||
strgeb r3, [r0, #-1]!
|
||||
ldrgtb r3, [r1, #-1]!
|
||||
strgtb r3, [r0, #-1]!
|
||||
ldmia sp!, {r0, pc}
|
||||
|
||||
/* erg - unaligned destination */
|
||||
.Lmemcpy_bdestul:
|
||||
cmp r12, #2
|
||||
|
||||
/* align destination with byte copies */
|
||||
ldrb r3, [r1, #-1]!
|
||||
strb r3, [r0, #-1]!
|
||||
ldrgeb r3, [r1, #-1]!
|
||||
strgeb r3, [r0, #-1]!
|
||||
ldrgtb r3, [r1, #-1]!
|
||||
strgtb r3, [r0, #-1]!
|
||||
subs r2, r2, r12
|
||||
blt .Lmemcpy_bl4 /* less than 4 bytes to go */
|
||||
ands r12, r1, #3
|
||||
beq .Lmemcpy_bt8 /* we have an aligned source */
|
||||
|
||||
/* erg - unaligned source */
|
||||
/* This is where it gets nasty ... */
|
||||
.Lmemcpy_bsrcul:
|
||||
bic r1, r1, #3
|
||||
ldr r3, [r1, #0]
|
||||
cmp r12, #2
|
||||
blt .Lmemcpy_bsrcul1
|
||||
beq .Lmemcpy_bsrcul2
|
||||
cmp r2, #0x0c
|
||||
blt .Lmemcpy_bsrcul3loop4
|
||||
sub r2, r2, #0x0c
|
||||
stmdb sp!, {r4, r5}
|
||||
|
||||
.Lmemcpy_bsrcul3loop16:
|
||||
#ifdef __ARMEB__
|
||||
mov lr, r3, lsr #8
|
||||
ldmdb r1!, {r3-r5, r12}
|
||||
orr lr, lr, r12, lsl #24
|
||||
mov r12, r12, lsr #8
|
||||
orr r12, r12, r5, lsl #24
|
||||
mov r5, r5, lsr #8
|
||||
orr r5, r5, r4, lsl #24
|
||||
mov r4, r4, lsr #8
|
||||
orr r4, r4, r3, lsl #24
|
||||
#else
|
||||
mov lr, r3, lsl #8
|
||||
ldmdb r1!, {r3-r5, r12}
|
||||
orr lr, lr, r12, lsr #24
|
||||
mov r12, r12, lsl #8
|
||||
orr r12, r12, r5, lsr #24
|
||||
mov r5, r5, lsl #8
|
||||
orr r5, r5, r4, lsr #24
|
||||
mov r4, r4, lsl #8
|
||||
orr r4, r4, r3, lsr #24
|
||||
#endif
|
||||
stmdb r0!, {r4, r5, r12, lr}
|
||||
subs r2, r2, #0x10
|
||||
bge .Lmemcpy_bsrcul3loop16
|
||||
ldmia sp!, {r4, r5}
|
||||
adds r2, r2, #0x0c
|
||||
blt .Lmemcpy_bsrcul3l4
|
||||
|
||||
.Lmemcpy_bsrcul3loop4:
|
||||
#ifdef __ARMEB__
|
||||
mov r12, r3, lsr #8
|
||||
ldr r3, [r1, #-4]!
|
||||
orr r12, r12, r3, lsl #24
|
||||
#else
|
||||
mov r12, r3, lsl #8
|
||||
ldr r3, [r1, #-4]!
|
||||
orr r12, r12, r3, lsr #24
|
||||
#endif
|
||||
str r12, [r0, #-4]!
|
||||
subs r2, r2, #4
|
||||
bge .Lmemcpy_bsrcul3loop4
|
||||
|
||||
.Lmemcpy_bsrcul3l4:
|
||||
add r1, r1, #3
|
||||
b .Lmemcpy_bl4
|
||||
|
||||
.Lmemcpy_bsrcul2:
|
||||
cmp r2, #0x0c
|
||||
blt .Lmemcpy_bsrcul2loop4
|
||||
sub r2, r2, #0x0c
|
||||
stmdb sp!, {r4, r5}
|
||||
|
||||
.Lmemcpy_bsrcul2loop16:
|
||||
#ifdef __ARMEB__
|
||||
mov lr, r3, lsr #16
|
||||
ldmdb r1!, {r3-r5, r12}
|
||||
orr lr, lr, r12, lsl #16
|
||||
mov r12, r12, lsr #16
|
||||
orr r12, r12, r5, lsl #16
|
||||
mov r5, r5, lsr #16
|
||||
orr r5, r5, r4, lsl #16
|
||||
mov r4, r4, lsr #16
|
||||
orr r4, r4, r3, lsl #16
|
||||
#else
|
||||
mov lr, r3, lsl #16
|
||||
ldmdb r1!, {r3-r5, r12}
|
||||
orr lr, lr, r12, lsr #16
|
||||
mov r12, r12, lsl #16
|
||||
orr r12, r12, r5, lsr #16
|
||||
mov r5, r5, lsl #16
|
||||
orr r5, r5, r4, lsr #16
|
||||
mov r4, r4, lsl #16
|
||||
orr r4, r4, r3, lsr #16
|
||||
#endif
|
||||
stmdb r0!, {r4, r5, r12, lr}
|
||||
subs r2, r2, #0x10
|
||||
bge .Lmemcpy_bsrcul2loop16
|
||||
ldmia sp!, {r4, r5}
|
||||
adds r2, r2, #0x0c
|
||||
blt .Lmemcpy_bsrcul2l4
|
||||
|
||||
.Lmemcpy_bsrcul2loop4:
|
||||
#ifdef __ARMEB__
|
||||
mov r12, r3, lsr #16
|
||||
ldr r3, [r1, #-4]!
|
||||
orr r12, r12, r3, lsl #16
|
||||
#else
|
||||
mov r12, r3, lsl #16
|
||||
ldr r3, [r1, #-4]!
|
||||
orr r12, r12, r3, lsr #16
|
||||
#endif
|
||||
str r12, [r0, #-4]!
|
||||
subs r2, r2, #4
|
||||
bge .Lmemcpy_bsrcul2loop4
|
||||
|
||||
.Lmemcpy_bsrcul2l4:
|
||||
add r1, r1, #2
|
||||
b .Lmemcpy_bl4
|
||||
|
||||
.Lmemcpy_bsrcul1:
|
||||
cmp r2, #0x0c
|
||||
blt .Lmemcpy_bsrcul1loop4
|
||||
sub r2, r2, #0x0c
|
||||
stmdb sp!, {r4, r5}
|
||||
|
||||
.Lmemcpy_bsrcul1loop32:
|
||||
#ifdef __ARMEB__
|
||||
mov lr, r3, lsr #24
|
||||
ldmdb r1!, {r3-r5, r12}
|
||||
orr lr, lr, r12, lsl #8
|
||||
mov r12, r12, lsr #24
|
||||
orr r12, r12, r5, lsl #8
|
||||
mov r5, r5, lsr #24
|
||||
orr r5, r5, r4, lsl #8
|
||||
mov r4, r4, lsr #24
|
||||
orr r4, r4, r3, lsl #8
|
||||
#else
|
||||
mov lr, r3, lsl #24
|
||||
ldmdb r1!, {r3-r5, r12}
|
||||
orr lr, lr, r12, lsr #8
|
||||
mov r12, r12, lsl #24
|
||||
orr r12, r12, r5, lsr #8
|
||||
mov r5, r5, lsl #24
|
||||
orr r5, r5, r4, lsr #8
|
||||
mov r4, r4, lsl #24
|
||||
orr r4, r4, r3, lsr #8
|
||||
#endif
|
||||
stmdb r0!, {r4, r5, r12, lr}
|
||||
subs r2, r2, #0x10
|
||||
bge .Lmemcpy_bsrcul1loop32
|
||||
ldmia sp!, {r4, r5}
|
||||
adds r2, r2, #0x0c
|
||||
blt .Lmemcpy_bsrcul1l4
|
||||
|
||||
.Lmemcpy_bsrcul1loop4:
|
||||
#ifdef __ARMEB__
|
||||
mov r12, r3, lsr #24
|
||||
ldr r3, [r1, #-4]!
|
||||
orr r12, r12, r3, lsl #8
|
||||
#else
|
||||
mov r12, r3, lsl #24
|
||||
ldr r3, [r1, #-4]!
|
||||
orr r12, r12, r3, lsr #8
|
||||
#endif
|
||||
str r12, [r0, #-4]!
|
||||
subs r2, r2, #4
|
||||
bge .Lmemcpy_bsrcul1loop4
|
||||
|
||||
.Lmemcpy_bsrcul1l4:
|
||||
add r1, r1, #1
|
||||
b .Lmemcpy_bl4
|
||||
|
||||
|
|
|
|||
|
|
@ -0,0 +1,273 @@
|
|||
/* $NetBSD: memcpy_arm.S,v 1.1 2003/10/13 19:59:24 scw Exp $ */
|
||||
|
||||
/*-
|
||||
* Copyright (c) 1997 The NetBSD Foundation, Inc.
|
||||
* All rights reserved.
|
||||
*
|
||||
* This code is derived from software contributed to The NetBSD Foundation
|
||||
* by Neil A. Carson and Mark Brinicombe
|
||||
*
|
||||
* 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 by the NetBSD
|
||||
* Foundation, Inc. and its contributors.
|
||||
* 4. Neither the name of The NetBSD Foundation nor the names of its
|
||||
* contributors may be used to endorse or promote products derived
|
||||
* from this software without specific prior written permission.
|
||||
*
|
||||
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
|
||||
* ``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 THE FOUNDATION OR CONTRIBUTORS
|
||||
* 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>
|
||||
|
||||
/*
|
||||
* This is one fun bit of code ...
|
||||
* Some easy listening music is suggested while trying to understand this
|
||||
* code e.g. Iron Maiden
|
||||
*
|
||||
* For anyone attempting to understand it :
|
||||
*
|
||||
* The core code is implemented here with simple stubs for memcpy().
|
||||
*
|
||||
* All local labels are prefixed with Lmemcpy_
|
||||
* Following the prefix a label starting f is used in the forward copy code
|
||||
* while a label using b is used in the backwards copy code
|
||||
* The source and destination addresses determine whether a forward or
|
||||
* backward copy is performed.
|
||||
* Separate bits of code are used to deal with the following situations
|
||||
* for both the forward and backwards copy.
|
||||
* unaligned source address
|
||||
* unaligned destination address
|
||||
* Separate copy routines are used to produce an optimised result for each
|
||||
* of these cases.
|
||||
* The copy code will use LDM/STM instructions to copy up to 32 bytes at
|
||||
* a time where possible.
|
||||
*
|
||||
* Note: r12 (aka ip) can be trashed during the function along with
|
||||
* r0-r3 although r0-r2 have defined uses i.e. src, dest, len through out.
|
||||
* Additional registers are preserved prior to use i.e. r4, r5 & lr
|
||||
*
|
||||
* Apologies for the state of the comments ;-)
|
||||
*/
|
||||
/* LINTSTUB: Func: void *memcpy(void *dst, const void *src, size_t len) */
|
||||
ENTRY(memcpy)
|
||||
/* save leaf functions having to store this away */
|
||||
stmdb sp!, {r0, lr} /* memcpy() returns dest addr */
|
||||
|
||||
subs r2, r2, #4
|
||||
blt .Lmemcpy_l4 /* less than 4 bytes */
|
||||
ands r12, r0, #3
|
||||
bne .Lmemcpy_destul /* oh unaligned destination addr */
|
||||
ands r12, r1, #3
|
||||
bne .Lmemcpy_srcul /* oh unaligned source addr */
|
||||
|
||||
.Lmemcpy_t8:
|
||||
/* We have aligned source and destination */
|
||||
subs r2, r2, #8
|
||||
blt .Lmemcpy_l12 /* less than 12 bytes (4 from above) */
|
||||
subs r2, r2, #0x14
|
||||
blt .Lmemcpy_l32 /* less than 32 bytes (12 from above) */
|
||||
stmdb sp!, {r4} /* borrow r4 */
|
||||
|
||||
/* blat 32 bytes at a time */
|
||||
/* XXX for really big copies perhaps we should use more registers */
|
||||
.Lmemcpy_loop32:
|
||||
ldmia r1!, {r3, r4, r12, lr}
|
||||
stmia r0!, {r3, r4, r12, lr}
|
||||
ldmia r1!, {r3, r4, r12, lr}
|
||||
stmia r0!, {r3, r4, r12, lr}
|
||||
subs r2, r2, #0x20
|
||||
bge .Lmemcpy_loop32
|
||||
|
||||
cmn r2, #0x10
|
||||
ldmgeia r1!, {r3, r4, r12, lr} /* blat a remaining 16 bytes */
|
||||
stmgeia r0!, {r3, r4, r12, lr}
|
||||
subge r2, r2, #0x10
|
||||
ldmia sp!, {r4} /* return r4 */
|
||||
|
||||
.Lmemcpy_l32:
|
||||
adds r2, r2, #0x14
|
||||
|
||||
/* blat 12 bytes at a time */
|
||||
.Lmemcpy_loop12:
|
||||
ldmgeia r1!, {r3, r12, lr}
|
||||
stmgeia r0!, {r3, r12, lr}
|
||||
subges r2, r2, #0x0c
|
||||
bge .Lmemcpy_loop12
|
||||
|
||||
.Lmemcpy_l12:
|
||||
adds r2, r2, #8
|
||||
blt .Lmemcpy_l4
|
||||
|
||||
subs r2, r2, #4
|
||||
ldrlt r3, [r1], #4
|
||||
strlt r3, [r0], #4
|
||||
ldmgeia r1!, {r3, r12}
|
||||
stmgeia r0!, {r3, r12}
|
||||
subge r2, r2, #4
|
||||
|
||||
.Lmemcpy_l4:
|
||||
/* less than 4 bytes to go */
|
||||
adds r2, r2, #4
|
||||
#ifdef __APCS_26_
|
||||
ldmeqia sp!, {r0, pc}^ /* done */
|
||||
#else
|
||||
ldmeqia sp!, {r0, pc} /* done */
|
||||
#endif
|
||||
/* copy the crud byte at a time */
|
||||
cmp r2, #2
|
||||
ldrb r3, [r1], #1
|
||||
strb r3, [r0], #1
|
||||
ldrgeb r3, [r1], #1
|
||||
strgeb r3, [r0], #1
|
||||
ldrgtb r3, [r1], #1
|
||||
strgtb r3, [r0], #1
|
||||
ldmia sp!, {r0, pc}
|
||||
|
||||
/* erg - unaligned destination */
|
||||
.Lmemcpy_destul:
|
||||
rsb r12, r12, #4
|
||||
cmp r12, #2
|
||||
|
||||
/* align destination with byte copies */
|
||||
ldrb r3, [r1], #1
|
||||
strb r3, [r0], #1
|
||||
ldrgeb r3, [r1], #1
|
||||
strgeb r3, [r0], #1
|
||||
ldrgtb r3, [r1], #1
|
||||
strgtb r3, [r0], #1
|
||||
subs r2, r2, r12
|
||||
blt .Lmemcpy_l4 /* less the 4 bytes */
|
||||
|
||||
ands r12, r1, #3
|
||||
beq .Lmemcpy_t8 /* we have an aligned source */
|
||||
|
||||
/* erg - unaligned source */
|
||||
/* This is where it gets nasty ... */
|
||||
.Lmemcpy_srcul:
|
||||
bic r1, r1, #3
|
||||
ldr lr, [r1], #4
|
||||
cmp r12, #2
|
||||
bgt .Lmemcpy_srcul3
|
||||
beq .Lmemcpy_srcul2
|
||||
cmp r2, #0x0c
|
||||
blt .Lmemcpy_srcul1loop4
|
||||
sub r2, r2, #0x0c
|
||||
stmdb sp!, {r4, r5}
|
||||
|
||||
.Lmemcpy_srcul1loop16:
|
||||
mov r3, lr, lsr #8
|
||||
ldmia r1!, {r4, r5, r12, lr}
|
||||
orr r3, r3, r4, lsl #24
|
||||
mov r4, r4, lsr #8
|
||||
orr r4, r4, r5, lsl #24
|
||||
mov r5, r5, lsr #8
|
||||
orr r5, r5, r12, lsl #24
|
||||
mov r12, r12, lsr #8
|
||||
orr r12, r12, lr, lsl #24
|
||||
stmia r0!, {r3-r5, r12}
|
||||
subs r2, r2, #0x10
|
||||
bge .Lmemcpy_srcul1loop16
|
||||
ldmia sp!, {r4, r5}
|
||||
adds r2, r2, #0x0c
|
||||
blt .Lmemcpy_srcul1l4
|
||||
|
||||
.Lmemcpy_srcul1loop4:
|
||||
mov r12, lr, lsr #8
|
||||
ldr lr, [r1], #4
|
||||
orr r12, r12, lr, lsl #24
|
||||
str r12, [r0], #4
|
||||
subs r2, r2, #4
|
||||
bge .Lmemcpy_srcul1loop4
|
||||
|
||||
.Lmemcpy_srcul1l4:
|
||||
sub r1, r1, #3
|
||||
b .Lmemcpy_l4
|
||||
|
||||
.Lmemcpy_srcul2:
|
||||
cmp r2, #0x0c
|
||||
blt .Lmemcpy_srcul2loop4
|
||||
sub r2, r2, #0x0c
|
||||
stmdb sp!, {r4, r5}
|
||||
|
||||
.Lmemcpy_srcul2loop16:
|
||||
mov r3, lr, lsr #16
|
||||
ldmia r1!, {r4, r5, r12, lr}
|
||||
orr r3, r3, r4, lsl #16
|
||||
mov r4, r4, lsr #16
|
||||
orr r4, r4, r5, lsl #16
|
||||
mov r5, r5, lsr #16
|
||||
orr r5, r5, r12, lsl #16
|
||||
mov r12, r12, lsr #16
|
||||
orr r12, r12, lr, lsl #16
|
||||
stmia r0!, {r3-r5, r12}
|
||||
subs r2, r2, #0x10
|
||||
bge .Lmemcpy_srcul2loop16
|
||||
ldmia sp!, {r4, r5}
|
||||
adds r2, r2, #0x0c
|
||||
blt .Lmemcpy_srcul2l4
|
||||
|
||||
.Lmemcpy_srcul2loop4:
|
||||
mov r12, lr, lsr #16
|
||||
ldr lr, [r1], #4
|
||||
orr r12, r12, lr, lsl #16
|
||||
str r12, [r0], #4
|
||||
subs r2, r2, #4
|
||||
bge .Lmemcpy_srcul2loop4
|
||||
|
||||
.Lmemcpy_srcul2l4:
|
||||
sub r1, r1, #2
|
||||
b .Lmemcpy_l4
|
||||
|
||||
.Lmemcpy_srcul3:
|
||||
cmp r2, #0x0c
|
||||
blt .Lmemcpy_srcul3loop4
|
||||
sub r2, r2, #0x0c
|
||||
stmdb sp!, {r4, r5}
|
||||
|
||||
.Lmemcpy_srcul3loop16:
|
||||
mov r3, lr, lsr #24
|
||||
ldmia r1!, {r4, r5, r12, lr}
|
||||
orr r3, r3, r4, lsl #8
|
||||
mov r4, r4, lsr #24
|
||||
orr r4, r4, r5, lsl #8
|
||||
mov r5, r5, lsr #24
|
||||
orr r5, r5, r12, lsl #8
|
||||
mov r12, r12, lsr #24
|
||||
orr r12, r12, lr, lsl #8
|
||||
stmia r0!, {r3-r5, r12}
|
||||
subs r2, r2, #0x10
|
||||
bge .Lmemcpy_srcul3loop16
|
||||
ldmia sp!, {r4, r5}
|
||||
adds r2, r2, #0x0c
|
||||
blt .Lmemcpy_srcul3l4
|
||||
|
||||
.Lmemcpy_srcul3loop4:
|
||||
mov r12, lr, lsr #24
|
||||
ldr lr, [r1], #4
|
||||
orr r12, r12, lr, lsl #8
|
||||
str r12, [r0], #4
|
||||
subs r2, r2, #4
|
||||
bge .Lmemcpy_srcul3loop4
|
||||
|
||||
.Lmemcpy_srcul3l4:
|
||||
sub r1, r1, #1
|
||||
b .Lmemcpy_l4
|
||||
File diff suppressed because it is too large
Load Diff
|
|
@ -1,5 +1,620 @@
|
|||
/* $NetBSD: memmove.S,v 1.2 2001/11/20 00:29:20 chris Exp $ */
|
||||
/* $NetBSD: memmove.S,v 1.3 2003/10/13 19:59:24 scw Exp $ */
|
||||
|
||||
/*-
|
||||
* Copyright (c) 1997 The NetBSD Foundation, Inc.
|
||||
* All rights reserved.
|
||||
*
|
||||
* This code is derived from software contributed to The NetBSD Foundation
|
||||
* by Neil A. Carson and Mark Brinicombe
|
||||
*
|
||||
* 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 by the NetBSD
|
||||
* Foundation, Inc. and its contributors.
|
||||
* 4. Neither the name of The NetBSD Foundation nor the names of its
|
||||
* contributors may be used to endorse or promote products derived
|
||||
* from this software without specific prior written permission.
|
||||
*
|
||||
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
|
||||
* ``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 THE FOUNDATION OR CONTRIBUTORS
|
||||
* 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>
|
||||
|
||||
/*
|
||||
* placeholder to keep the make system happy, memove is actually in memcpy.S
|
||||
* This is one fun bit of code ...
|
||||
* Some easy listening music is suggested while trying to understand this
|
||||
* code e.g. Iron Maiden
|
||||
*
|
||||
* For anyone attempting to understand it :
|
||||
*
|
||||
* The core code is implemented here with simple stubs for memmove()
|
||||
*
|
||||
* All local labels are prefixed with Lmemmove_
|
||||
* Following the prefix a label starting f is used in the forward copy code
|
||||
* while a label using b is used in the backwards copy code
|
||||
* The source and destination addresses determine whether a forward or
|
||||
* backward copy is performed.
|
||||
* Separate bits of code are used to deal with the following situations
|
||||
* for both the forward and backwards copy.
|
||||
* unaligned source address
|
||||
* unaligned destination address
|
||||
* Separate copy routines are used to produce an optimised result for each
|
||||
* of these cases.
|
||||
* The copy code will use LDM/STM instructions to copy up to 32 bytes at
|
||||
* a time where possible.
|
||||
*
|
||||
* Note: r12 (aka ip) can be trashed during the function along with
|
||||
* r0-r3 although r0-r2 have defined uses i.e. src, dest, len through out.
|
||||
* Additional registers are preserved prior to use i.e. r4, r5 & lr
|
||||
*
|
||||
* Apologies for the state of the comments ;-)
|
||||
*/
|
||||
|
||||
ENTRY(memmove)
|
||||
#ifdef __XSCALE__
|
||||
/*
|
||||
* The XSCALE version of memcpy() is *way* faster than
|
||||
* this code, so see if we can use it.
|
||||
*/
|
||||
/* Do the buffers overlap? */
|
||||
cmp r0, r1
|
||||
moveq pc, lr /* Bail now if src/dst are the same */
|
||||
subcc r3, r0, r1 /* if (dst > src) r3 = dst - src */
|
||||
subcs r3, r1, r0 /* if (src > dsr) r3 = src - dst */
|
||||
cmp r3, r2 /* if (r3 < len) we have an overlap */
|
||||
bcc _C_LABEL(memcpy)
|
||||
#endif
|
||||
|
||||
/* Determine copy direction */
|
||||
cmp r1, r0
|
||||
|
||||
moveq r0, #0 /* Quick abort for len=0 */
|
||||
moveq pc, lr
|
||||
|
||||
/* save leaf functions having to store this away */
|
||||
stmdb sp!, {r0, lr} /* memmove() returns dest addr */
|
||||
|
||||
bcc .Lmemmove_backwards
|
||||
|
||||
/* start of forwards copy */
|
||||
subs r2, r2, #4
|
||||
blt .Lmemmove_fl4 /* less than 4 bytes */
|
||||
ands r12, r0, #3
|
||||
bne .Lmemmove_fdestul /* oh unaligned destination addr */
|
||||
ands r12, r1, #3
|
||||
bne .Lmemmove_fsrcul /* oh unaligned source addr */
|
||||
|
||||
.Lmemmove_ft8:
|
||||
/* We have aligned source and destination */
|
||||
subs r2, r2, #8
|
||||
blt .Lmemmove_fl12 /* less than 12 bytes (4 from above) */
|
||||
subs r2, r2, #0x14
|
||||
blt .Lmemmove_fl32 /* less than 32 bytes (12 from above) */
|
||||
stmdb sp!, {r4} /* borrow r4 */
|
||||
|
||||
/* blat 32 bytes at a time */
|
||||
/* XXX for really big copies perhaps we should use more registers */
|
||||
.Lmemmove_floop32:
|
||||
ldmia r1!, {r3, r4, r12, lr}
|
||||
stmia r0!, {r3, r4, r12, lr}
|
||||
ldmia r1!, {r3, r4, r12, lr}
|
||||
stmia r0!, {r3, r4, r12, lr}
|
||||
subs r2, r2, #0x20
|
||||
bge .Lmemmove_floop32
|
||||
|
||||
cmn r2, #0x10
|
||||
ldmgeia r1!, {r3, r4, r12, lr} /* blat a remaining 16 bytes */
|
||||
stmgeia r0!, {r3, r4, r12, lr}
|
||||
subge r2, r2, #0x10
|
||||
ldmia sp!, {r4} /* return r4 */
|
||||
|
||||
.Lmemmove_fl32:
|
||||
adds r2, r2, #0x14
|
||||
|
||||
/* blat 12 bytes at a time */
|
||||
.Lmemmove_floop12:
|
||||
ldmgeia r1!, {r3, r12, lr}
|
||||
stmgeia r0!, {r3, r12, lr}
|
||||
subges r2, r2, #0x0c
|
||||
bge .Lmemmove_floop12
|
||||
|
||||
.Lmemmove_fl12:
|
||||
adds r2, r2, #8
|
||||
blt .Lmemmove_fl4
|
||||
|
||||
subs r2, r2, #4
|
||||
ldrlt r3, [r1], #4
|
||||
strlt r3, [r0], #4
|
||||
ldmgeia r1!, {r3, r12}
|
||||
stmgeia r0!, {r3, r12}
|
||||
subge r2, r2, #4
|
||||
|
||||
.Lmemmove_fl4:
|
||||
/* less than 4 bytes to go */
|
||||
adds r2, r2, #4
|
||||
#ifdef __APCS_26_
|
||||
ldmeqia sp!, {r0, pc}^ /* done */
|
||||
#else
|
||||
ldmeqia sp!, {r0, pc} /* done */
|
||||
#endif
|
||||
/* copy the crud byte at a time */
|
||||
cmp r2, #2
|
||||
ldrb r3, [r1], #1
|
||||
strb r3, [r0], #1
|
||||
ldrgeb r3, [r1], #1
|
||||
strgeb r3, [r0], #1
|
||||
ldrgtb r3, [r1], #1
|
||||
strgtb r3, [r0], #1
|
||||
ldmia sp!, {r0, pc}
|
||||
|
||||
/* erg - unaligned destination */
|
||||
.Lmemmove_fdestul:
|
||||
rsb r12, r12, #4
|
||||
cmp r12, #2
|
||||
|
||||
/* align destination with byte copies */
|
||||
ldrb r3, [r1], #1
|
||||
strb r3, [r0], #1
|
||||
ldrgeb r3, [r1], #1
|
||||
strgeb r3, [r0], #1
|
||||
ldrgtb r3, [r1], #1
|
||||
strgtb r3, [r0], #1
|
||||
subs r2, r2, r12
|
||||
blt .Lmemmove_fl4 /* less the 4 bytes */
|
||||
|
||||
ands r12, r1, #3
|
||||
beq .Lmemmove_ft8 /* we have an aligned source */
|
||||
|
||||
/* erg - unaligned source */
|
||||
/* This is where it gets nasty ... */
|
||||
.Lmemmove_fsrcul:
|
||||
bic r1, r1, #3
|
||||
ldr lr, [r1], #4
|
||||
cmp r12, #2
|
||||
bgt .Lmemmove_fsrcul3
|
||||
beq .Lmemmove_fsrcul2
|
||||
cmp r2, #0x0c
|
||||
blt .Lmemmove_fsrcul1loop4
|
||||
sub r2, r2, #0x0c
|
||||
stmdb sp!, {r4, r5}
|
||||
|
||||
.Lmemmove_fsrcul1loop16:
|
||||
#ifdef __ARMEB__
|
||||
mov r3, lr, lsl #8
|
||||
#else
|
||||
mov r3, lr, lsr #8
|
||||
#endif
|
||||
ldmia r1!, {r4, r5, r12, lr}
|
||||
#ifdef __ARMEB__
|
||||
orr r3, r3, r4, lsl #24
|
||||
mov r4, r4, lsr #8
|
||||
orr r4, r4, r5, lsl #24
|
||||
mov r5, r5, lsr #8
|
||||
orr r5, r5, r12, lsl #24
|
||||
mov r12, r12, lsr #8
|
||||
orr r12, r12, lr, lsl #24
|
||||
#else
|
||||
orr r3, r3, r4, lsr #24
|
||||
mov r4, r4, lsl #8
|
||||
orr r4, r4, r5, lsr #24
|
||||
mov r5, r5, lsl #8
|
||||
orr r5, r5, r12, lsr #24
|
||||
mov r12, r12, lsl #8
|
||||
orr r12, r12, lr, lsr #24
|
||||
#endif
|
||||
stmia r0!, {r3-r5, r12}
|
||||
subs r2, r2, #0x10
|
||||
bge .Lmemmove_fsrcul1loop16
|
||||
ldmia sp!, {r4, r5}
|
||||
adds r2, r2, #0x0c
|
||||
blt .Lmemmove_fsrcul1l4
|
||||
|
||||
.Lmemmove_fsrcul1loop4:
|
||||
#ifdef __ARMEB__
|
||||
mov r12, lr, lsl #8
|
||||
#else
|
||||
mov r12, lr, lsr #8
|
||||
#endif
|
||||
ldr lr, [r1], #4
|
||||
#ifdef __ARMEB__
|
||||
orr r12, r12, lr, lsr #24
|
||||
#else
|
||||
orr r12, r12, lr, lsl #24
|
||||
#endif
|
||||
str r12, [r0], #4
|
||||
subs r2, r2, #4
|
||||
bge .Lmemmove_fsrcul1loop4
|
||||
|
||||
.Lmemmove_fsrcul1l4:
|
||||
sub r1, r1, #3
|
||||
b .Lmemmove_fl4
|
||||
|
||||
.Lmemmove_fsrcul2:
|
||||
cmp r2, #0x0c
|
||||
blt .Lmemmove_fsrcul2loop4
|
||||
sub r2, r2, #0x0c
|
||||
stmdb sp!, {r4, r5}
|
||||
|
||||
.Lmemmove_fsrcul2loop16:
|
||||
#ifdef __ARMEB__
|
||||
mov r3, lr, lsl #16
|
||||
#else
|
||||
mov r3, lr, lsr #16
|
||||
#endif
|
||||
ldmia r1!, {r4, r5, r12, lr}
|
||||
#ifdef __ARMEB__
|
||||
orr r3, r3, r4, lsr #16
|
||||
mov r4, r4, lsl #16
|
||||
orr r4, r4, r5, lsr #16
|
||||
mov r5, r5, lsl #16
|
||||
orr r5, r5, r12, lsr #16
|
||||
mov r12, r12, lsl #16
|
||||
orr r12, r12, lr, lsr #16
|
||||
#else
|
||||
orr r3, r3, r4, lsl #16
|
||||
mov r4, r4, lsr #16
|
||||
orr r4, r4, r5, lsl #16
|
||||
mov r5, r5, lsr #16
|
||||
orr r5, r5, r12, lsl #16
|
||||
mov r12, r12, lsr #16
|
||||
orr r12, r12, lr, lsl #16
|
||||
#endif
|
||||
stmia r0!, {r3-r5, r12}
|
||||
subs r2, r2, #0x10
|
||||
bge .Lmemmove_fsrcul2loop16
|
||||
ldmia sp!, {r4, r5}
|
||||
adds r2, r2, #0x0c
|
||||
blt .Lmemmove_fsrcul2l4
|
||||
|
||||
.Lmemmove_fsrcul2loop4:
|
||||
#ifdef __ARMEB__
|
||||
mov r12, lr, lsl #16
|
||||
#else
|
||||
mov r12, lr, lsr #16
|
||||
#endif
|
||||
ldr lr, [r1], #4
|
||||
#ifdef __ARMEB__
|
||||
orr r12, r12, lr, lsr #16
|
||||
#else
|
||||
orr r12, r12, lr, lsl #16
|
||||
#endif
|
||||
str r12, [r0], #4
|
||||
subs r2, r2, #4
|
||||
bge .Lmemmove_fsrcul2loop4
|
||||
|
||||
.Lmemmove_fsrcul2l4:
|
||||
sub r1, r1, #2
|
||||
b .Lmemmove_fl4
|
||||
|
||||
.Lmemmove_fsrcul3:
|
||||
cmp r2, #0x0c
|
||||
blt .Lmemmove_fsrcul3loop4
|
||||
sub r2, r2, #0x0c
|
||||
stmdb sp!, {r4, r5}
|
||||
|
||||
.Lmemmove_fsrcul3loop16:
|
||||
#ifdef __ARMEB__
|
||||
mov r3, lr, lsl #24
|
||||
#else
|
||||
mov r3, lr, lsr #24
|
||||
#endif
|
||||
ldmia r1!, {r4, r5, r12, lr}
|
||||
#ifdef __ARMEB__
|
||||
orr r3, r3, r4, lsr #8
|
||||
mov r4, r4, lsl #24
|
||||
orr r4, r4, r5, lsr #8
|
||||
mov r5, r5, lsl #24
|
||||
orr r5, r5, r12, lsr #8
|
||||
mov r12, r12, lsl #24
|
||||
orr r12, r12, lr, lsr #8
|
||||
#else
|
||||
orr r3, r3, r4, lsl #8
|
||||
mov r4, r4, lsr #24
|
||||
orr r4, r4, r5, lsl #8
|
||||
mov r5, r5, lsr #24
|
||||
orr r5, r5, r12, lsl #8
|
||||
mov r12, r12, lsr #24
|
||||
orr r12, r12, lr, lsl #8
|
||||
#endif
|
||||
stmia r0!, {r3-r5, r12}
|
||||
subs r2, r2, #0x10
|
||||
bge .Lmemmove_fsrcul3loop16
|
||||
ldmia sp!, {r4, r5}
|
||||
adds r2, r2, #0x0c
|
||||
blt .Lmemmove_fsrcul3l4
|
||||
|
||||
.Lmemmove_fsrcul3loop4:
|
||||
#ifdef __ARMEB__
|
||||
mov r12, lr, lsl #24
|
||||
#else
|
||||
mov r12, lr, lsr #24
|
||||
#endif
|
||||
ldr lr, [r1], #4
|
||||
#ifdef __ARMEB__
|
||||
orr r12, r12, lr, lsr #8
|
||||
#else
|
||||
orr r12, r12, lr, lsl #8
|
||||
#endif
|
||||
str r12, [r0], #4
|
||||
subs r2, r2, #4
|
||||
bge .Lmemmove_fsrcul3loop4
|
||||
|
||||
.Lmemmove_fsrcul3l4:
|
||||
sub r1, r1, #1
|
||||
b .Lmemmove_fl4
|
||||
|
||||
.Lmemmove_backwards:
|
||||
add r1, r1, r2
|
||||
add r0, r0, r2
|
||||
subs r2, r2, #4
|
||||
blt .Lmemmove_bl4 /* less than 4 bytes */
|
||||
ands r12, r0, #3
|
||||
bne .Lmemmove_bdestul /* oh unaligned destination addr */
|
||||
ands r12, r1, #3
|
||||
bne .Lmemmove_bsrcul /* oh unaligned source addr */
|
||||
|
||||
.Lmemmove_bt8:
|
||||
/* We have aligned source and destination */
|
||||
subs r2, r2, #8
|
||||
blt .Lmemmove_bl12 /* less than 12 bytes (4 from above) */
|
||||
stmdb sp!, {r4}
|
||||
subs r2, r2, #0x14 /* less than 32 bytes (12 from above) */
|
||||
blt .Lmemmove_bl32
|
||||
|
||||
/* blat 32 bytes at a time */
|
||||
/* XXX for really big copies perhaps we should use more registers */
|
||||
.Lmemmove_bloop32:
|
||||
ldmdb r1!, {r3, r4, r12, lr}
|
||||
stmdb r0!, {r3, r4, r12, lr}
|
||||
ldmdb r1!, {r3, r4, r12, lr}
|
||||
stmdb r0!, {r3, r4, r12, lr}
|
||||
subs r2, r2, #0x20
|
||||
bge .Lmemmove_bloop32
|
||||
|
||||
.Lmemmove_bl32:
|
||||
cmn r2, #0x10
|
||||
ldmgedb r1!, {r3, r4, r12, lr} /* blat a remaining 16 bytes */
|
||||
stmgedb r0!, {r3, r4, r12, lr}
|
||||
subge r2, r2, #0x10
|
||||
adds r2, r2, #0x14
|
||||
ldmgedb r1!, {r3, r12, lr} /* blat a remaining 12 bytes */
|
||||
stmgedb r0!, {r3, r12, lr}
|
||||
subge r2, r2, #0x0c
|
||||
ldmia sp!, {r4}
|
||||
|
||||
.Lmemmove_bl12:
|
||||
adds r2, r2, #8
|
||||
blt .Lmemmove_bl4
|
||||
subs r2, r2, #4
|
||||
ldrlt r3, [r1, #-4]!
|
||||
strlt r3, [r0, #-4]!
|
||||
ldmgedb r1!, {r3, r12}
|
||||
stmgedb r0!, {r3, r12}
|
||||
subge r2, r2, #4
|
||||
|
||||
.Lmemmove_bl4:
|
||||
/* less than 4 bytes to go */
|
||||
adds r2, r2, #4
|
||||
ldmeqia sp!, {r0, pc}
|
||||
|
||||
/* copy the crud byte at a time */
|
||||
cmp r2, #2
|
||||
ldrb r3, [r1, #-1]!
|
||||
strb r3, [r0, #-1]!
|
||||
ldrgeb r3, [r1, #-1]!
|
||||
strgeb r3, [r0, #-1]!
|
||||
ldrgtb r3, [r1, #-1]!
|
||||
strgtb r3, [r0, #-1]!
|
||||
ldmia sp!, {r0, pc}
|
||||
|
||||
/* erg - unaligned destination */
|
||||
.Lmemmove_bdestul:
|
||||
cmp r12, #2
|
||||
|
||||
/* align destination with byte copies */
|
||||
ldrb r3, [r1, #-1]!
|
||||
strb r3, [r0, #-1]!
|
||||
ldrgeb r3, [r1, #-1]!
|
||||
strgeb r3, [r0, #-1]!
|
||||
ldrgtb r3, [r1, #-1]!
|
||||
strgtb r3, [r0, #-1]!
|
||||
subs r2, r2, r12
|
||||
blt .Lmemmove_bl4 /* less than 4 bytes to go */
|
||||
ands r12, r1, #3
|
||||
beq .Lmemmove_bt8 /* we have an aligned source */
|
||||
|
||||
/* erg - unaligned source */
|
||||
/* This is where it gets nasty ... */
|
||||
.Lmemmove_bsrcul:
|
||||
bic r1, r1, #3
|
||||
ldr r3, [r1, #0]
|
||||
cmp r12, #2
|
||||
blt .Lmemmove_bsrcul1
|
||||
beq .Lmemmove_bsrcul2
|
||||
cmp r2, #0x0c
|
||||
blt .Lmemmove_bsrcul3loop4
|
||||
sub r2, r2, #0x0c
|
||||
stmdb sp!, {r4, r5}
|
||||
|
||||
.Lmemmove_bsrcul3loop16:
|
||||
#ifdef __ARMEB__
|
||||
mov lr, r3, lsr #8
|
||||
#else
|
||||
mov lr, r3, lsl #8
|
||||
#endif
|
||||
ldmdb r1!, {r3-r5, r12}
|
||||
#ifdef __ARMEB__
|
||||
orr lr, lr, r12, lsl #24
|
||||
mov r12, r12, lsr #8
|
||||
orr r12, r12, r5, lsl #24
|
||||
mov r5, r5, lsr #8
|
||||
orr r5, r5, r4, lsl #24
|
||||
mov r4, r4, lsr #8
|
||||
orr r4, r4, r3, lsl #24
|
||||
#else
|
||||
orr lr, lr, r12, lsr #24
|
||||
mov r12, r12, lsl #8
|
||||
orr r12, r12, r5, lsr #24
|
||||
mov r5, r5, lsl #8
|
||||
orr r5, r5, r4, lsr #24
|
||||
mov r4, r4, lsl #8
|
||||
orr r4, r4, r3, lsr #24
|
||||
#endif
|
||||
stmdb r0!, {r4, r5, r12, lr}
|
||||
subs r2, r2, #0x10
|
||||
bge .Lmemmove_bsrcul3loop16
|
||||
ldmia sp!, {r4, r5}
|
||||
adds r2, r2, #0x0c
|
||||
blt .Lmemmove_bsrcul3l4
|
||||
|
||||
.Lmemmove_bsrcul3loop4:
|
||||
#ifdef __ARMEB__
|
||||
mov r12, r3, lsr #8
|
||||
#else
|
||||
mov r12, r3, lsl #8
|
||||
#endif
|
||||
ldr r3, [r1, #-4]!
|
||||
#ifdef __ARMEB__
|
||||
orr r12, r12, r3, lsl #24
|
||||
#else
|
||||
orr r12, r12, r3, lsr #24
|
||||
#endif
|
||||
str r12, [r0, #-4]!
|
||||
subs r2, r2, #4
|
||||
bge .Lmemmove_bsrcul3loop4
|
||||
|
||||
.Lmemmove_bsrcul3l4:
|
||||
add r1, r1, #3
|
||||
b .Lmemmove_bl4
|
||||
|
||||
.Lmemmove_bsrcul2:
|
||||
cmp r2, #0x0c
|
||||
blt .Lmemmove_bsrcul2loop4
|
||||
sub r2, r2, #0x0c
|
||||
stmdb sp!, {r4, r5}
|
||||
|
||||
.Lmemmove_bsrcul2loop16:
|
||||
#ifdef __ARMEB__
|
||||
mov lr, r3, lsr #16
|
||||
#else
|
||||
mov lr, r3, lsl #16
|
||||
#endif
|
||||
ldmdb r1!, {r3-r5, r12}
|
||||
#ifdef __ARMEB__
|
||||
orr lr, lr, r12, lsl #16
|
||||
mov r12, r12, lsr #16
|
||||
orr r12, r12, r5, lsl #16
|
||||
mov r5, r5, lsr #16
|
||||
orr r5, r5, r4, lsl #16
|
||||
mov r4, r4, lsr #16
|
||||
orr r4, r4, r3, lsl #16
|
||||
#else
|
||||
orr lr, lr, r12, lsr #16
|
||||
mov r12, r12, lsl #16
|
||||
orr r12, r12, r5, lsr #16
|
||||
mov r5, r5, lsl #16
|
||||
orr r5, r5, r4, lsr #16
|
||||
mov r4, r4, lsl #16
|
||||
orr r4, r4, r3, lsr #16
|
||||
#endif
|
||||
stmdb r0!, {r4, r5, r12, lr}
|
||||
subs r2, r2, #0x10
|
||||
bge .Lmemmove_bsrcul2loop16
|
||||
ldmia sp!, {r4, r5}
|
||||
adds r2, r2, #0x0c
|
||||
blt .Lmemmove_bsrcul2l4
|
||||
|
||||
.Lmemmove_bsrcul2loop4:
|
||||
#ifdef __ARMEB__
|
||||
mov r12, r3, lsr #16
|
||||
#else
|
||||
mov r12, r3, lsl #16
|
||||
#endif
|
||||
ldr r3, [r1, #-4]!
|
||||
#ifdef __ARMEB__
|
||||
orr r12, r12, r3, lsl #16
|
||||
#else
|
||||
orr r12, r12, r3, lsr #16
|
||||
#endif
|
||||
str r12, [r0, #-4]!
|
||||
subs r2, r2, #4
|
||||
bge .Lmemmove_bsrcul2loop4
|
||||
|
||||
.Lmemmove_bsrcul2l4:
|
||||
add r1, r1, #2
|
||||
b .Lmemmove_bl4
|
||||
|
||||
.Lmemmove_bsrcul1:
|
||||
cmp r2, #0x0c
|
||||
blt .Lmemmove_bsrcul1loop4
|
||||
sub r2, r2, #0x0c
|
||||
stmdb sp!, {r4, r5}
|
||||
|
||||
.Lmemmove_bsrcul1loop32:
|
||||
#ifdef __ARMEB__
|
||||
mov lr, r3, lsr #24
|
||||
#else
|
||||
mov lr, r3, lsl #24
|
||||
#endif
|
||||
ldmdb r1!, {r3-r5, r12}
|
||||
#ifdef __ARMEB__
|
||||
orr lr, lr, r12, lsl #8
|
||||
mov r12, r12, lsr #24
|
||||
orr r12, r12, r5, lsl #8
|
||||
mov r5, r5, lsr #24
|
||||
orr r5, r5, r4, lsl #8
|
||||
mov r4, r4, lsr #24
|
||||
orr r4, r4, r3, lsl #8
|
||||
#else
|
||||
orr lr, lr, r12, lsr #8
|
||||
mov r12, r12, lsl #24
|
||||
orr r12, r12, r5, lsr #8
|
||||
mov r5, r5, lsl #24
|
||||
orr r5, r5, r4, lsr #8
|
||||
mov r4, r4, lsl #24
|
||||
orr r4, r4, r3, lsr #8
|
||||
#endif
|
||||
stmdb r0!, {r4, r5, r12, lr}
|
||||
subs r2, r2, #0x10
|
||||
bge .Lmemmove_bsrcul1loop32
|
||||
ldmia sp!, {r4, r5}
|
||||
adds r2, r2, #0x0c
|
||||
blt .Lmemmove_bsrcul1l4
|
||||
|
||||
.Lmemmove_bsrcul1loop4:
|
||||
#ifdef __ARMEB__
|
||||
mov r12, r3, lsr #24
|
||||
#else
|
||||
mov r12, r3, lsl #24
|
||||
#endif
|
||||
ldr r3, [r1, #-4]!
|
||||
#ifdef __ARMEB__
|
||||
orr r12, r12, r3, lsl #8
|
||||
#else
|
||||
orr r12, r12, r3, lsr #8
|
||||
#endif
|
||||
str r12, [r0, #-4]!
|
||||
subs r2, r2, #4
|
||||
bge .Lmemmove_bsrcul1loop4
|
||||
|
||||
.Lmemmove_bsrcul1l4:
|
||||
add r1, r1, #1
|
||||
b .Lmemmove_bl4
|
||||
|
|
|
|||
|
|
@ -1,5 +1,39 @@
|
|||
/* $NetBSD: memset.S,v 1.3 2003/04/05 23:27:15 bjh21 Exp $ */
|
||||
/* $NetBSD: memset.S,v 1.4 2003/10/13 19:59:24 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) 1995 Mark Brinicombe.
|
||||
* All rights reserved.
|
||||
|
|
@ -35,7 +69,7 @@
|
|||
#include <machine/asm.h>
|
||||
|
||||
/*
|
||||
* Sets a block of memory to the specified value
|
||||
* memset: Sets a block of memory to the specified value
|
||||
*
|
||||
* On entry:
|
||||
* r0 - dest address
|
||||
|
|
@ -45,82 +79,157 @@
|
|||
* On exit:
|
||||
* r0 - dest address
|
||||
*/
|
||||
|
||||
#ifdef _BZERO
|
||||
/* LINTSTUB: Func: void bzero(void *, size_t) */
|
||||
ENTRY(bzero)
|
||||
mov r3, #0x00
|
||||
#else
|
||||
/* LINTSTUB: Func: void *memset(void *, int, size_t) */
|
||||
ENTRY(memset)
|
||||
stmfd sp!, {r0} /* Remember address for return value */
|
||||
and r1, r1, #0x000000ff /* We write bytes */
|
||||
|
||||
cmp r2, #0x00000004 /* Do we have less than 4 bytes */
|
||||
and r3, r1, #0xff /* We deal with bytes */
|
||||
mov r1, r2
|
||||
#endif
|
||||
cmp r1, #0x04 /* Do we have less than 4 bytes */
|
||||
mov ip, r0
|
||||
blt .Lmemset_lessthanfour
|
||||
|
||||
/* Ok first we will word align the address */
|
||||
ands r2, ip, #0x03 /* Get the bottom two bits */
|
||||
bne .Lmemset_wordunaligned /* The address is not word aligned */
|
||||
|
||||
ands r3, r0, #0x00000003 /* Get the bottom two bits */
|
||||
beq .Lmemset_addraligned /* The address is word aligned */
|
||||
/* We are now word aligned */
|
||||
.Lmemset_wordaligned:
|
||||
#ifndef _BZERO
|
||||
orr r3, r3, r3, lsl #8 /* Extend value to 16-bits */
|
||||
#endif
|
||||
#ifdef __XSCALE__
|
||||
tst ip, #0x04 /* Quad-align for Xscale */
|
||||
#else
|
||||
cmp r1, #0x10
|
||||
#endif
|
||||
#ifndef _BZERO
|
||||
orr r3, r3, r3, lsl #16 /* Extend value to 32-bits */
|
||||
#endif
|
||||
#ifdef __XSCALE__
|
||||
subne r1, r1, #0x04 /* Quad-align if necessary */
|
||||
strne r3, [ip], #0x04
|
||||
cmp r1, #0x10
|
||||
#endif
|
||||
blt .Lmemset_loop4 /* If less than 16 then use words */
|
||||
mov r2, r3 /* Duplicate data */
|
||||
cmp r1, #0x80 /* If < 128 then skip the big loop */
|
||||
blt .Lmemset_loop32
|
||||
|
||||
rsb r3, r3, #0x00000004
|
||||
sub r2, r2, r3
|
||||
cmp r3, #0x00000002
|
||||
strb r1, [r0], #0x0001 /* Set 1 byte */
|
||||
strgeb r1, [r0], #0x0001 /* Set another byte */
|
||||
strgtb r1, [r0], #0x0001 /* and a third */
|
||||
/* Do 128 bytes at a time */
|
||||
.Lmemset_loop128:
|
||||
subs r1, r1, #0x80
|
||||
#ifdef __XSCALE__
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
#else
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
#endif
|
||||
bgt .Lmemset_loop128
|
||||
moveq pc, lr /* Zero length so just exit */
|
||||
|
||||
cmp r2, #0x00000004
|
||||
blt .Lmemset_lessthanfour
|
||||
add r1, r1, #0x80 /* Adjust for extra sub */
|
||||
|
||||
/* Now we must be word aligned */
|
||||
/* Do 32 bytes at a time */
|
||||
.Lmemset_loop32:
|
||||
subs r1, r1, #0x20
|
||||
#ifdef __XSCALE__
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
#else
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
#endif
|
||||
bgt .Lmemset_loop32
|
||||
moveq pc, lr /* Zero length so just exit */
|
||||
|
||||
.Lmemset_addraligned:
|
||||
adds r1, r1, #0x10 /* Partially adjust for extra sub */
|
||||
|
||||
orr r3, r1, r1, lsl #8 /* Repeat the byte into a word */
|
||||
orr r3, r3, r3, lsl #16
|
||||
/* Deal with 16 bytes or more */
|
||||
#ifdef __XSCALE__
|
||||
strged r2, [ip], #0x08
|
||||
strged r2, [ip], #0x08
|
||||
#else
|
||||
stmgeia ip!, {r2-r3}
|
||||
stmgeia ip!, {r2-r3}
|
||||
#endif
|
||||
moveq pc, lr /* Zero length so just exit */
|
||||
|
||||
/* We know we have at least 4 bytes ... */
|
||||
|
||||
cmp r2, #0x00000020 /* If less than 32 then use words */
|
||||
blt .Lmemset_lessthan32
|
||||
|
||||
/* We have at least 32 so lets use quad words */
|
||||
|
||||
stmfd sp!, {r4-r6} /* Store registers */
|
||||
mov r4, r3 /* Duplicate data */
|
||||
mov r5, r3
|
||||
mov r6, r3
|
||||
|
||||
.Lmemset_loop16:
|
||||
stmia r0!, {r3-r6} /* Store 16 bytes */
|
||||
sub r2, r2, #0x00000010 /* Adjust count */
|
||||
cmp r2, #0x00000010 /* Still got at least 16 bytes ? */
|
||||
bgt .Lmemset_loop16
|
||||
|
||||
ldmfd sp!, {r4-r6} /* Restore registers */
|
||||
|
||||
/* Do we need to set some words as well ? */
|
||||
|
||||
cmp r2, #0x00000004
|
||||
blt .Lmemset_lessthanfour
|
||||
|
||||
/* Have either less than 16 or less than 32 depending on route taken */
|
||||
|
||||
.Lmemset_lessthan32:
|
||||
addlt r1, r1, #0x10 /* Possibly adjust for extra sub */
|
||||
|
||||
/* We have at least 4 bytes so copy as words */
|
||||
|
||||
.Lmemset_loop4:
|
||||
str r3, [r0], #0x0004
|
||||
sub r2, r2, #0x0004
|
||||
cmp r2, #0x00000004
|
||||
bge .Lmemset_loop4
|
||||
subs r1, r1, #0x04
|
||||
strge r3, [ip], #0x04
|
||||
bgt .Lmemset_loop4
|
||||
moveq pc, lr /* Zero length so just exit */
|
||||
|
||||
#ifdef __XSCALE__
|
||||
/* Compensate for 64-bit alignment check */
|
||||
adds r1, r1, #0x04
|
||||
moveq pc, lr
|
||||
cmp r1, #2
|
||||
#else
|
||||
cmp r1, #-2
|
||||
#endif
|
||||
|
||||
strb r3, [ip], #0x01 /* Set 1 byte */
|
||||
strgeb r3, [ip], #0x01 /* Set another byte */
|
||||
strgtb r3, [ip] /* and a third */
|
||||
mov pc, lr /* Exit */
|
||||
|
||||
.Lmemset_wordunaligned:
|
||||
rsb r2, r2, #0x004
|
||||
strb r3, [ip], #0x01 /* Set 1 byte */
|
||||
cmp r2, #0x02
|
||||
strgeb r3, [ip], #0x01 /* Set another byte */
|
||||
sub r1, r1, r2
|
||||
strgtb r3, [ip], #0x01 /* and a third */
|
||||
cmp r1, #0x04 /* More than 4 bytes left? */
|
||||
bge .Lmemset_wordaligned /* Yup */
|
||||
|
||||
.Lmemset_lessthanfour:
|
||||
cmp r2, #0x00000000
|
||||
ldmeqfd sp!, {r0}
|
||||
cmp r1, #0x00
|
||||
moveq pc, lr /* Zero length so exit */
|
||||
|
||||
cmp r2, #0x00000002
|
||||
strb r1, [r0], #0x0001 /* Set 1 byte */
|
||||
strgeb r1, [r0], #0x0001 /* Set another byte */
|
||||
strgtb r1, [r0], #0x0001 /* and a third */
|
||||
|
||||
ldmfd sp!, {r0}
|
||||
strb r3, [ip], #0x01 /* Set 1 byte */
|
||||
cmp r1, #0x02
|
||||
strgeb r3, [ip], #0x01 /* Set another byte */
|
||||
strgtb r3, [ip] /* and a third */
|
||||
mov pc, lr /* Exit */
|
||||
|
|
|
|||
Loading…
Reference in New Issue