1/*
2 *  linux/arch/arm/lib/copypage.S
3 *
4 *  Copyright (C) 1995-1999 Russell King
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
9 *
10 *  ASM optimised string functions
11 */
12#include <linux/linkage.h>
13#include <asm/assembler.h>
14#include <asm/asm-offsets.h>
15#include <asm/cache.h>
16
17#define COPY_COUNT (PAGE_SZ / (2 * L1_CACHE_BYTES) PLD( -1 ))
18
19		.text
20		.align	5
21/*
22 * StrongARM optimised copy_page routine
23 * now 1.78bytes/cycle, was 1.60 bytes/cycle (50MHz bus -> 89MB/s)
24 * Note that we probably achieve closer to the 100MB/s target with
25 * the core clock switching.
26 */
27ENTRY(copy_page)
28		stmfd	sp!, {r4, lr}			@	2
29	PLD(	pld	[r1, #0]		)
30	PLD(	pld	[r1, #L1_CACHE_BYTES]		)
31		mov	r2, #COPY_COUNT			@	1
32		ldmia	r1!, {r3, r4, ip, lr}		@	4+1
331:	PLD(	pld	[r1, #2 * L1_CACHE_BYTES])
34	PLD(	pld	[r1, #3 * L1_CACHE_BYTES])
352:
36	.rept	(2 * L1_CACHE_BYTES / 16 - 1)
37		stmia	r0!, {r3, r4, ip, lr}		@	4
38		ldmia	r1!, {r3, r4, ip, lr}		@	4
39	.endr
40		subs	r2, r2, #1			@	1
41		stmia	r0!, {r3, r4, ip, lr}		@	4
42		ldmgtia	r1!, {r3, r4, ip, lr}		@	4
43		bgt	1b				@	1
44	PLD(	ldmeqia r1!, {r3, r4, ip, lr}	)
45	PLD(	beq	2b			)
46		ldmfd	sp!, {r4, pc}			@	3
47ENDPROC(copy_page)
48