1/*
  2 *  linux/arch/arm/lib/memset.S
  3 *
  4 *  Copyright (C) 1995-2000 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/unwind.h>
 15
 16	.text
 17	.align	5
 18
 19ENTRY(mmioset)
 20ENTRY(memset)
 21UNWIND( .fnstart         )
 22	ands	r3, r0, #3		@ 1 unaligned?
 23	mov	ip, r0			@ preserve r0 as return value
 24	bne	6f			@ 1
 25/*
 26 * we know that the pointer in ip is aligned to a word boundary.
 27 */
 281:	orr	r1, r1, r1, lsl #8
 29	orr	r1, r1, r1, lsl #16
 30	mov	r3, r1
 31	cmp	r2, #16
 32	blt	4f
 33
 34#if ! CALGN(1)+0
 35
 36/*
 37 * We need 2 extra registers for this loop - use r8 and the LR
 38 */
 39	stmfd	sp!, {r8, lr}
 40UNWIND( .fnend              )
 41UNWIND( .fnstart            )
 42UNWIND( .save {r8, lr}      )
 43	mov	r8, r1
 44	mov	lr, r1
 45
 462:	subs	r2, r2, #64
 47	stmgeia	ip!, {r1, r3, r8, lr}	@ 64 bytes at a time.
 48	stmgeia	ip!, {r1, r3, r8, lr}
 49	stmgeia	ip!, {r1, r3, r8, lr}
 50	stmgeia	ip!, {r1, r3, r8, lr}
 51	bgt	2b
 52	ldmeqfd	sp!, {r8, pc}		@ Now <64 bytes to go.
 53/*
 54 * No need to correct the count; we're only testing bits from now on
 55 */
 56	tst	r2, #32
 57	stmneia	ip!, {r1, r3, r8, lr}
 58	stmneia	ip!, {r1, r3, r8, lr}
 59	tst	r2, #16
 60	stmneia	ip!, {r1, r3, r8, lr}
 61	ldmfd	sp!, {r8, lr}
 62UNWIND( .fnend              )
 63
 64#else
 65
 66/*
 67 * This version aligns the destination pointer in order to write
 68 * whole cache lines at once.
 69 */
 70
 71	stmfd	sp!, {r4-r8, lr}
 72UNWIND( .fnend                 )
 73UNWIND( .fnstart               )
 74UNWIND( .save {r4-r8, lr}      )
 75	mov	r4, r1
 76	mov	r5, r1
 77	mov	r6, r1
 78	mov	r7, r1
 79	mov	r8, r1
 80	mov	lr, r1
 81
 82	cmp	r2, #96
 83	tstgt	ip, #31
 84	ble	3f
 85
 86	and	r8, ip, #31
 87	rsb	r8, r8, #32
 88	sub	r2, r2, r8
 89	movs	r8, r8, lsl #(32 - 4)
 90	stmcsia	ip!, {r4, r5, r6, r7}
 91	stmmiia	ip!, {r4, r5}
 92	tst	r8, #(1 << 30)
 93	mov	r8, r1
 94	strne	r1, [ip], #4
 95
 963:	subs	r2, r2, #64
 97	stmgeia	ip!, {r1, r3-r8, lr}
 98	stmgeia	ip!, {r1, r3-r8, lr}
 99	bgt	3b
100	ldmeqfd	sp!, {r4-r8, pc}
101
102	tst	r2, #32
103	stmneia	ip!, {r1, r3-r8, lr}
104	tst	r2, #16
105	stmneia	ip!, {r4-r7}
106	ldmfd	sp!, {r4-r8, lr}
107UNWIND( .fnend                 )
108
109#endif
110
111UNWIND( .fnstart            )
1124:	tst	r2, #8
113	stmneia	ip!, {r1, r3}
114	tst	r2, #4
115	strne	r1, [ip], #4
116/*
117 * When we get here, we've got less than 4 bytes to zero.  We
118 * may have an unaligned pointer as well.
119 */
1205:	tst	r2, #2
121	strneb	r1, [ip], #1
122	strneb	r1, [ip], #1
123	tst	r2, #1
124	strneb	r1, [ip], #1
125	ret	lr
126
1276:	subs	r2, r2, #4		@ 1 do we have enough
128	blt	5b			@ 1 bytes to align with?
129	cmp	r3, #2			@ 1
130	strltb	r1, [ip], #1		@ 1
131	strleb	r1, [ip], #1		@ 1
132	strb	r1, [ip], #1		@ 1
133	add	r2, r2, r3		@ 1 (r2 = r2 - (4 - r3))
134	b	1b
135UNWIND( .fnend   )
136ENDPROC(memset)
137ENDPROC(mmioset)

  1/*
  2 *  linux/arch/arm/lib/memset.S
  3 *
  4 *  Copyright (C) 1995-2000 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
 15	.text
 16	.align	5
 17
 
 18ENTRY(memset)
 
 19	ands	r3, r0, #3		@ 1 unaligned?
 20	mov	ip, r0			@ preserve r0 as return value
 21	bne	6f			@ 1
 22/*
 23 * we know that the pointer in ip is aligned to a word boundary.
 24 */
 251:	orr	r1, r1, r1, lsl #8
 26	orr	r1, r1, r1, lsl #16
 27	mov	r3, r1
 28	cmp	r2, #16
 29	blt	4f
 30
 31#if ! CALGN(1)+0
 32
 33/*
 34 * We need 2 extra registers for this loop - use r8 and the LR
 35 */
 36	stmfd	sp!, {r8, lr}
 
 
 
 37	mov	r8, r1
 38	mov	lr, r1
 39
 402:	subs	r2, r2, #64
 41	stmgeia	ip!, {r1, r3, r8, lr}	@ 64 bytes at a time.
 42	stmgeia	ip!, {r1, r3, r8, lr}
 43	stmgeia	ip!, {r1, r3, r8, lr}
 44	stmgeia	ip!, {r1, r3, r8, lr}
 45	bgt	2b
 46	ldmeqfd	sp!, {r8, pc}		@ Now <64 bytes to go.
 47/*
 48 * No need to correct the count; we're only testing bits from now on
 49 */
 50	tst	r2, #32
 51	stmneia	ip!, {r1, r3, r8, lr}
 52	stmneia	ip!, {r1, r3, r8, lr}
 53	tst	r2, #16
 54	stmneia	ip!, {r1, r3, r8, lr}
 55	ldmfd	sp!, {r8, lr}
 
 56
 57#else
 58
 59/*
 60 * This version aligns the destination pointer in order to write
 61 * whole cache lines at once.
 62 */
 63
 64	stmfd	sp!, {r4-r8, lr}
 
 
 
 65	mov	r4, r1
 66	mov	r5, r1
 67	mov	r6, r1
 68	mov	r7, r1
 69	mov	r8, r1
 70	mov	lr, r1
 71
 72	cmp	r2, #96
 73	tstgt	ip, #31
 74	ble	3f
 75
 76	and	r8, ip, #31
 77	rsb	r8, r8, #32
 78	sub	r2, r2, r8
 79	movs	r8, r8, lsl #(32 - 4)
 80	stmcsia	ip!, {r4, r5, r6, r7}
 81	stmmiia	ip!, {r4, r5}
 82	tst	r8, #(1 << 30)
 83	mov	r8, r1
 84	strne	r1, [ip], #4
 85
 863:	subs	r2, r2, #64
 87	stmgeia	ip!, {r1, r3-r8, lr}
 88	stmgeia	ip!, {r1, r3-r8, lr}
 89	bgt	3b
 90	ldmeqfd	sp!, {r4-r8, pc}
 91
 92	tst	r2, #32
 93	stmneia	ip!, {r1, r3-r8, lr}
 94	tst	r2, #16
 95	stmneia	ip!, {r4-r7}
 96	ldmfd	sp!, {r4-r8, lr}
 
 97
 98#endif
 99
 
1004:	tst	r2, #8
101	stmneia	ip!, {r1, r3}
102	tst	r2, #4
103	strne	r1, [ip], #4
104/*
105 * When we get here, we've got less than 4 bytes to zero.  We
106 * may have an unaligned pointer as well.
107 */
1085:	tst	r2, #2
109	strneb	r1, [ip], #1
110	strneb	r1, [ip], #1
111	tst	r2, #1
112	strneb	r1, [ip], #1
113	mov	pc, lr
114
1156:	subs	r2, r2, #4		@ 1 do we have enough
116	blt	5b			@ 1 bytes to align with?
117	cmp	r3, #2			@ 1
118	strltb	r1, [ip], #1		@ 1
119	strleb	r1, [ip], #1		@ 1
120	strb	r1, [ip], #1		@ 1
121	add	r2, r2, r3		@ 1 (r2 = r2 - (4 - r3))
122	b	1b
 
123ENDPROC(memset)